EfficientNetV2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CodeTestEfficientNetV2.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "uXLQLk2diUIG"
      ]
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cgeLhcnaDLMV",
        "outputId": "a80556f7-442f-4bc0-fd7d-c5e3eab63384"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "O1YIrS7xrHBb"
      },
      "source": [
        "# Libs\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PUd1h1y_BaVL",
        "outputId": "e113c43f-d898-4059-d7f7-f84405c73cc2"
      },
      "source": [
        "!pip install imutils"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: imutils in /usr/local/lib/python3.7/dist-packages (0.5.4)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y6cxwjlHBtxh"
      },
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "from scipy.spatial import distance as dist\n",
        "import matplotlib.pyplot as plt\n",
        "import os\n",
        "import cv2\n",
        "import seaborn as sns\n",
        "from tqdm import tqdm \n",
        "from sklearn.utils import shuffle\n",
        "#from sklearn import decomposition\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix, plot_confusion_matrix\n",
        "# import tensorflow as tf\n",
        "import keras\n",
        "#from keras.applications.vgg16 import VGG16 \n",
        "from keras.preprocessing import image\n",
        "#from keras.applications.vgg16 import preprocess_input\n",
        "from keras.models import Sequential, Model \n",
        "#from keras.initializers import he_normal\n",
        "# from keras.layers import Lambda, SeparableConv2D, BatchNormalization, Dropout, MaxPooling2D, Input, Dense, Conv2D, Activation, Flatten \n",
        "from keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint\n",
        "#from keras.preprocessing.image import ImageDataGenerator\n",
        "import imutils\n",
        "import itertools"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nyGILex0B_6t"
      },
      "source": [
        "class_names = ['EOSINOPHIL', 'LYMPHOCYTE', 'MONOCYTE', 'NEUTROPHIL']\n",
        "nb_classes = len(class_names)"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aiyjbXKoJvz-"
      },
      "source": [
        "# Load Image\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zlabUXQt12EY"
      },
      "source": [
        "def findEdges(image):\n",
        "    # find edges in image\n",
        "    gray = cv2.GaussianBlur(image, (1, 1), 0)\n",
        "    edged = cv2.Canny(gray, 100, 400)\n",
        "    edged = cv2.dilate(edged, None, iterations=1)\n",
        "    edged = cv2.erode(edged, None, iterations=1)\n",
        "    return edged"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "78hEVLTgB7FI"
      },
      "source": [
        "def getBoxes(contours, orig):\n",
        "    boxes = []\n",
        "    centers = []\n",
        "    for contour in contours:\n",
        "        box = cv2.minAreaRect(contour)\n",
        "        box = cv2.cv.BoxPoints(box) if imutils.is_cv2() else cv2.boxPoints(box)\n",
        "        box = np.array(box, dtype=\"int\")\n",
        "        (tl, tr, br, bl) = box\n",
        "        if (dist.euclidean(tl, bl)) > 0 and (dist.euclidean(tl, tr)) > 0:\n",
        "            boxes.append(box)\n",
        "    return boxes"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "av0VWaH-B4b2"
      },
      "source": [
        "def getImgContours(edged):\n",
        "    contours = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)\n",
        "    contours = imutils.grab_contours(contours)\n",
        "    contours = sorted(contours, key=lambda x: cv2.contourArea(x))\n",
        "    return contours"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AEmFUr24wiEF"
      },
      "source": [
        "IMAGE_SIZE = 128\n",
        "EPOCHS = 50"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "selB3yS1CDK9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "397c9d72-d3a1-4abb-f801-9ec0d21be4a5"
      },
      "source": [
        "\n",
        "datasets = ['/content/drive/MyDrive/BME2/dataset2-master/images/TRAIN', '/content/drive/MyDrive/BME2/dataset2-master/images/TEST']\n",
        "images = []\n",
        "labels = []\n",
        "  #lặp qua các folder trong thư mục datasets\n",
        "count =0\n",
        "for dataset in datasets:\n",
        "\n",
        "    for folder in os.listdir(dataset):\n",
        "\n",
        "        if folder in ['EOSINOPHIL']: label = 0\n",
        "        elif folder in ['LYMPHOCYTE']: label = 1\n",
        "        elif folder in ['MONOCYTE']: label = 2\n",
        "        elif folder in ['NEUTROPHIL']: label = 3\n",
        "\n",
        "        for file in tqdm(os.listdir(os.path.join(dataset, folder))):\n",
        "\n",
        "            img_path = os.path.join(os.path.join(dataset, folder), file)\n",
        "\n",
        "            # Open \n",
        "            image = cv2.imread(img_path)\n",
        "            image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
        "            \n",
        "            # add padding to the image to better detect cell at the edge\n",
        "            image = cv2.copyMakeBorder(image,10,10,10,10,cv2.BORDER_CONSTANT,value=[198, 203, 208])\n",
        "            \n",
        "            #thresholding the image to get the target cell\n",
        "            image1 = cv2.inRange(image,(80, 80, 180),(180, 170, 245))\n",
        "            \n",
        "            # openning errosion then dilation\n",
        "            kernel = np.ones((3, 3), np.uint8)\n",
        "            kernel1 = np.ones((5, 5), np.uint8)\n",
        "            img_erosion = cv2.erode(image1, kernel, iterations=2)\n",
        "            image1 = cv2.dilate(img_erosion, kernel1, iterations=5)\n",
        "            \n",
        "            #detecting the blood cell\n",
        "            edgedImage = findEdges(image1)\n",
        "            edgedContours = getImgContours(edgedImage)\n",
        "            edgedBoxes =  getBoxes(edgedContours, image.copy())\n",
        "            if len(edgedBoxes)==0:\n",
        "                count +=1\n",
        "                continue\n",
        "            # get the large box and get its cordinate\n",
        "            last = edgedBoxes[-1]\n",
        "            max_x = int(max(last[:,0]))\n",
        "            min_x = int( min(last[:,0]))\n",
        "            max_y = int(max(last[:,1]))\n",
        "            min_y = int(min(last[:,1]))\n",
        "            \n",
        "            # draw the contour and fill it \n",
        "            mask = np.zeros_like(image)\n",
        "            cv2.drawContours(mask, edgedContours, len(edgedContours)-1, (255,255,255), -1) \n",
        "            \n",
        "            # any pixel but the pixels inside the contour is zero\n",
        "            image[mask==0] = 0\n",
        "            \n",
        "            # extract th blood cell\n",
        "            image = image[min_y:max_y, min_x:max_x]\n",
        "\n",
        "            if (np.size(image)==0):\n",
        "                count +=1\n",
        "                continue\n",
        "            # resize th image\n",
        "            image = cv2.resize(image, (IMAGE_SIZE,IMAGE_SIZE))\n",
        "\n",
        "            # Append the image and its corresponding label to the output\n",
        "            images.append(image)\n",
        "            labels.append(label)\n",
        "\n",
        "images = np.array(images, dtype = 'float32')\n",
        "labels = np.array(labels, dtype = 'int32')\n",
        "\n",
        "# return images, labels\n"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 2499/2499 [15:04<00:00,  2.76it/s]\n",
            "100%|██████████| 2478/2478 [14:37<00:00,  2.82it/s]\n",
            "100%|██████████| 2497/2497 [13:56<00:00,  2.98it/s]\n",
            "100%|██████████| 2483/2483 [14:10<00:00,  2.92it/s]\n",
            "100%|██████████| 624/624 [03:39<00:00,  2.85it/s]\n",
            "100%|██████████| 623/623 [03:41<00:00,  2.82it/s]\n",
            "100%|██████████| 620/620 [03:46<00:00,  2.74it/s]\n",
            "100%|██████████| 620/620 [03:51<00:00,  2.67it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "55R8B3ez_uET",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d53c2577-99de-4c3d-92b8-9c139e29cc31"
      },
      "source": [
        "print(len(images))\n",
        "print(len(labels))"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "12146\n",
            "12146\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BT_X3EqwCGAJ"
      },
      "source": [
        "images, labels = shuffle(images, labels, random_state=10)\n",
        "\n",
        "train_images, test_images, train_labels, test_labels = train_test_split(images, labels, test_size = 0.2)\n",
        "test_images, val_images, test_labels, val_labels = train_test_split(test_images, test_labels, test_size = 0.5)"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EjKoPrJo63Vi",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "900dea5f-99d3-44b6-8494-a02b725670b6"
      },
      "source": [
        "print(len(train_images))\n",
        "print(len(train_labels))\n",
        "print(len(val_images))\n",
        "print(len(val_labels))\n",
        "print(len(test_images))\n",
        "print(len(test_labels))"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "9716\n",
            "9716\n",
            "1215\n",
            "1215\n",
            "1215\n",
            "1215\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ew10yySbCItt",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "309c9a9c-9a64-4ee8-b199-3947eed4c994"
      },
      "source": [
        "\n",
        "print(\"Training images: {}\".format(train_images.shape))\n",
        "print(\"Training labels: {}\".format(train_labels.shape))\n",
        "print(\"Validation images: {}\".format(val_images.shape))\n",
        "print(\"Validation labels: {}\".format(val_labels.shape))\n",
        "print(\"Test images: {}\".format(test_images.shape))\n",
        "print(\"Test labels: {}\".format(test_labels.shape))"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Training images: (9716, 128, 128, 3)\n",
            "Training labels: (9716,)\n",
            "Validation images: (1215, 128, 128, 3)\n",
            "Validation labels: (1215,)\n",
            "Test images: (1215, 128, 128, 3)\n",
            "Test labels: (1215,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 314
        },
        "id": "Y4LMlziEcY4Z",
        "outputId": "8fa985b5-c486-4a05-be28-8f74d664ec12"
      },
      "source": [
        "#hiển thị train image  \n",
        "print(len(train_images))\n",
        "index = np.random.randint(train_images.shape[0])\n",
        "plt.figure()\n",
        "plt.imshow(train_images[index])\n",
        "plt.title('Image #{}: '.format(index) + class_names[train_labels[index]])\n",
        "plt.show()\n",
        "    "
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "9716\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 266
        },
        "id": "I5mcWrGWFVJT",
        "outputId": "c0e12761-3df4-497b-d61a-f5bd9fec312c"
      },
      "source": [
        "# Ground truth\n",
        "# Hiển thị ảnh từ train data\n",
        "plt.subplot(121)\n",
        "index = np.random.randint(train_images.shape[0])\n",
        "plt.imshow(train_images[index,:,:], cmap='gray')\n",
        "plt.title(\"Ground Truth : {}\".format(train_labels[index]))\n",
        "\n",
        "# Hiển thị ảnh từ test data\n",
        "plt.subplot(122)\n",
        "index = np.random.randint(test_images.shape[0])\n",
        "plt.imshow(test_images[index,:,:], cmap='gray')\n",
        "plt.title(\"Ground Truth : {}\".format(test_labels[index]))\n",
        "\n"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 1.0, 'Ground Truth : 0')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KmIBxUMjCLVb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 320
        },
        "outputId": "14ad5f9e-8195-454d-848c-3bc400642f97"
      },
      "source": [
        "_, train_counts = np.unique(train_labels, return_counts = True)\n",
        "_, val_counts = np.unique(val_labels, return_counts = True)\n",
        "_, test_counts = np.unique(test_labels, return_counts = True)\n",
        "\n",
        "pd.DataFrame({'train': train_counts, \"val\": val_counts, \"test\": test_counts}, index = class_names).plot.bar()\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4o_c4FymCRZW",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 264
        },
        "outputId": "31d1ff93-d370-46c6-9cc1-bc772b816627"
      },
      "source": [
        "plt.pie(train_counts,\n",
        "        explode=(0, 0, 0, 0) , \n",
        "        labels=class_names,\n",
        "        autopct='%1.1f%%')\n",
        "plt.axis('equal')\n",
        "plt.title('Proportion of each observed category')\n",
        "plt.show()"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 314
        },
        "id": "QQHW_3yMH-gb",
        "outputId": "90b4cc3b-927a-4e68-b98c-2fab0d94ea39"
      },
      "source": [
        "def display_random_image (class_names, images, labels):\n",
        "    print(len(images))\n",
        "    index = np.random.randint(images.shape[0])\n",
        "    plt.figure()\n",
        "    plt.imshow(images[index])\n",
        "    plt.title('Image #{}: '.format(index) + class_names[labels[index]])\n",
        "    plt.show()\n",
        "    \n",
        "display_random_image (class_names, train_images, train_labels)"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "9716\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQEAAAEICAYAAABf40E1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAZjklEQVR4nO3de5hU9Z3n8ffHRi5eQBEGEFDwMiqa2eiyakZXXUlUkKiT+Ph4mYSoM+y4Jt7YMRIna9Qx4piM0TWjw0giyRiVeEVMTFR0fIwLgomKgkjjhcugNCJysQVavvvH+bVUtwU0XV2X5nxez9NPV/3Oqfp9+3TVp37nUucoIjCz/Nqp2gWYWXU5BMxyziFglnMOAbOccwiY5ZxDwCznHAJmOZfbEJD0jqQvV7uOtpI0TNIL6fZ1ki5pNf1vJNVLWivpCUl7t5p+hKTn0vT3JV1aMO16SXMkNUn6QavHDZA0VdJ/SgpJQ7az7nckNaZ+m39uL5g+SNI9kj6QtE7Si5JGt3qO0yW9LGm1pBWSpksamqb9QNK/F8wb6W/ZqaDtHyXdXXC/m6QbJS1KtS2Q9PeSVDDPs5I+SfWukPSQpAFp2t2S/rFVjUNS311az9N6Wq3JbQh0Qv8VmF1w+4/NEySdAPwQOB3oDbwN3FswvQ/wBPCvwF7AAcDvC567HrgSeLxIv5vSY79eQu1fjYjdCn6+nerqDTwPbAAOBfoAtwC/knRmmucA4BfAOKAXMBT4KfDpVvrbGzh7K9N/DYwARgG7A98AxgK3tprv2xGxG/DnwB6pth2OQwCQ9C1Jf5B0i6RVkt6S9JepfbGk5ZLGFMx/qqQ/pU+mxUU+Pb8p6d306fb9wlGHpJ0kXSVpYZo+Jb0ZtmU48FK6fTjwcsG00cCvI+L1iNgAXA8cJ2n/NP0K4HcRcU9ErI+INRExr/nBETE5In4LrGndaUS8HxH/AsxqQ43b63JgLXBhRLwXEY0RcS9wA/Dj9Mn8ReDtiHg6Mmsi4sGIWLSV5/0n4Npin7ySRgAnAV+PiNcioikiZgB/DVycQqeFiFgJPAgcVuofXIscApsdBbxK9kn5K+A+4L+RfWr+NXC7pN3SvOuAb5J9OpwKXCTpDMiG7cC/AOcBA8g+vQYW9PMd4AzgeLJPrA/JPtmKkvSkpFXAxcD/lbQa6AcskfTbwlmL3G5+0R4NrJT0Qgq0xyTt06alsg0p0Ka18+FfAR6MiE2t2qcA+5B9Av8RODgF9P8o+B9szUPAauBbW+hzZkQsLmyMiJnAErIRQgtpJPV14E9t6LvTcQhs9nZE/DwiPgXuBwYD16VPzt+TDVkPAIiIZyNiTkRsiohXyYbex6fnORN4LCKeT5/K/wco/ILG3wFXR8SSiFgP/AA4c0vrixHxFeBI4OWI6AlMAK6KiD0iYmSa7QngLEl/IalHQZ+7pOmDgDHApWRvrharC6WIiAkRMXobsz2SRljNP3+b2vsAy4rM39zWJyLeAk4gC9IpwIq0vr21MAjg+8D3JXVtNW1LfTb326fg/m0pgF9J064omPa/C/8msg+QTskhsNn7BbcbIRsKt2rbDUDSUZKekdQg6SOyN3bzi2dv4LNPmYj4GPig4Hn2BR4uePHMI1u/7de6IEnfLngRHppuXw/8Q3r8n6U+ngKuIRuyvpN+1pB9sjXX/nBEzIqIT4Brgb+U1Gs7lk8pzkih1fzzb6l9BdloqbUBBdOJiBkRcVZE9AX+O3AccPXWOoyI35D9/f+z1aQt9dnc74qC+5ekegdGxHkR0VAw7UeFfxPwF1urp5Y5BNrnV8BUYHBE9ALuZPMQfBnZJy8A6ZN5r4LHLgZGtnpTdI+Ipa07iYjb0wvsP4ATyQJkaUT0So9bXjDvTyPiwIjoRxYGXYDX0uRXaTkaqZWvjj4FfK1wS35yFtlyerP1AyJiFtlwvy3r51cD32PziKi5z6MkDS6cUdJRZKO/6W2ufgfhEGif3YGVEfGJpCOBcwumPQB8NW1Y7Eo23C9cX78TuEHSvgCS+ko6fRv9fZFsNHAEBXsFmknqLukwZfYBJgK3RsSHaZafA38l6YuSdiYbKj8fER+lx+8sqTvZ66FLer66wucHuqW73dL9jnAL2TaTSZL6p37PIXvz/n1EhKRjJf1t86hH0sHAacCMbT15RDxLFoRjCtqeAp4GHpR0qKQ6SUcD/w7cERELOuhvK6Zb+hubf2ri/VcTRXRC/wu4TtIasvXvKc0TIuJ1so1/95GNCtYCy4H1aZZbyUYRv0+Pn0G2UbKo9Kb+IK1WHMHmPQSFupONTtYCLwL/j+yN3lzTdLJPxMdTLQfQMrj+jWyVofkN2Ei226xZY3pugDfS/eb6vtdqA2Uxj6nlcQIPp7o+AI5N9c8lW226AvhGRNyfHruK7E0/R9Jasu0fD5PtAWiLfyDbbVro68Az6bnWkgXAJLL/WzmtJVt2zT8nlrm/NpFPKlJeaQPWKuDAiHi72vWYteaRQBlI+qqkXSTtCvwImEO2sc6s5jgEyuN04D/Tz4HA2eEhl9Wosq0OSDqFbP23DrgrIiaUpSMzK0lZQiBtWX6T7OisJWSHnJ4TEXM7vDMzK0m5vtV0JFCfjvZC0n1kQ+SiISCpKkPl3r17M3To0Gp0bVYxCxcuZNWqVQAr0gFXLZQrBAZScNQc2WigxW4wSWPJvrlVNSNHjuSXv/wlkrY9s1knFBF87Wtf45FHHgF4t9g8Vft+c0RMJDuopWojATMr396BpWSHYDYblNrMrMaUayQwCzhQ2dlflpKd4OHcrT+kvHbeeWe6devWoq1Hjx5Vqsascnr06MFuu+3G2rVri04v5y7CUcBPyHYR/iwibtjKvGVfHbjgggsYN25ci7ZevXoxcODALTzCrPOLCJYsWcKaNWs49NBDX4qI4a3nKds2gfRVzt+U6/m3V58+fRg2bFi1yzCrKEkMHjx4q/P4iEGznKvJs5+W6qSTTuKoo1p+Me/YY4+tUjVmtW2HDIGRI0dy2WWXVbsMs07BqwNmOecQMMs5h4BZzjkEzHLOIWCWczvE3oGePXty4403stde2Zm9Dz/88CpXZNZ57BAh0L17d0aPHs0++3TIlbXMcsWrA2Y55xAwyzmHgFnOOQTMcq7Th0DXrl3p0aOHzxNo1k6dfu/Addddx6mnnkq/fp+7sreZtUGnD4F9992Xww5ry1WqzayYTr86YGalcQiY5ZxDwCznHAJmOecQMMs5h4BZznW6XYQ9e/Zk5MiRdO3aFYAhQ4ZUtyCzTq7ThUD//v2588476dWrF4CPFDQrUbtXByQNlvSMpLmSXpd0aWrvLelJSQvS7z07rtwW/TsAzDpAKdsEmoBxETEMOBq4WNIw4Crg6Yg4EHg63TezGtXuEIiIZRHxx3R7DTAPGAicDkxOs00Gzii1SDMrnw7ZJiBpCHA4MBPoFxHL0qT3gKLf7JE0FhjbEf2bWfuVvItQ0m7Ag8BlEbG6cFpk1z0vetnxiJgYEcOLXSrZzCqnpBCQtDNZANwTEQ+l5vclDUjTBwDLSyvRzMqplL0DAiYB8yLinwsmTQXGpNtjgEfbX56ZlVsp2wSOAb4BzJH0cmr7HjABmCLpQuBd4KzSSjSzcmp3CETE88CWdtSPaO/zbs3+++/PIYccQl1dXTme3iyXOs0Rg126dOH222/n+OOPp3v37tUux2yH0WlCALIrDfXo0aPaZZjtUGoyBIodDuzDhM3Ko+ZC4KKLLmLUqFGfa5fkE4qalUHNhcAXvvAFRo8eXe0yzHLDJxUxyzmHgFnOOQTMcs4hYJZzNbNhsEuXLuy6666fnTvQzCqjZkLghBNO4LbbbqN///7VLsUsV2omBHbffXcOPvhgHxBkVmHeJmCWcw4Bs5xzCJjlnEPALOccAmY55xAwyzmHgFnOOQTMcs4hYJZzDgGznHMImOVcTYRAr169OPLII6tdhlkuKbtmaHUNHz48Zs2a5S8PmZWRpJeKXQC4I65KXCfpT5KmpftDJc2UVC/pfkltOkGAA8CsOjpideBSYF7B/ZuAWyLiAOBD4MIO6MPMyqTUS5MPAk4F7kr3BZwIPJBmmQycUUofZlZepY4EfgJcCWxK9/cCVkVEU7q/BBhY7IGSxkqaLWl2Q0NDiWWYWXu1OwQkjQaWR8RL7Xl8REyMiOERMbxv377tLcPMSlTK6cWOAU6TNAroDvQEbgX2kNQljQYGAUtLL9PMyqXdI4GIGB8RgyJiCHA2MD0izgOeAc5Ms40BHi25SjMrm3IcLPRd4ApJ9WTbCCaVoQ8z6yAdcrbhiHgWeDbdfgvw4X9mnURNHDZsZtXjEDDLOYeAWc45BMxyziFglnMOAbOccwiY5ZxDwCznHAJmOecQMMs5h4BZzjkEzHLOIWCWcw4Bs5xzCJjlnEPALOccAmY55xAwyzmHgFnOOQTMcs4hYJZzHXK2YTOrTRHBunXr2Lhx4xbn8UjAbAd35ZVXctRRR21xukPAbAe3bNkyFixYsMXpDgGznCtpm4CkPYC7gMOAAC4A5gP3A0OAd4CzIuLDkqo0s21qampi0qRJLFq0qEX73Llzt/q4UjcM3go8ERFnSuoK7AJ8D3g6IiZIugq4iuz6hGZWRps2beLuu+9mxowZ2/W4dq8OSOoFHEe64GhEbIiIVcDpwOQ022TgjPb2YWblV8pIYCjQAPxc0n8BXgIuBfpFxLI0z3tAv2IPljQWGAuwzz77lFCGWT6sWrWKxYsXb3H6xo0b+fjjj7f7eUsJgS7AEcB3ImKmpFvJhv6fiYiQFMUeHBETgYkAw4cPLzqPmW325JNPcv755291nsbGxu1+3lJCYAmwJCJmpvsPkIXA+5IGRMQySQOA5SX0YWZJU1MT69at6/Dnbfc2gYh4D1gs6aDUNAKYC0wFxqS2McCjJVVolnMR8dlPOZS6d+A7wD1pz8BbwPlkwTJF0oXAu8BZJfZhlmsNDQ2MGzeOefPmleX5SwqBiHgZGF5k0ohSntfMNmtsbOSJJ55gxYoVZXl+HzFolnP+FqFZDVqwYAFz5swBstWBDRs2lK0vh4BZDXrssccYN25cRfry6oBZznkkYFZBjY2NNDU1bXO+9evXV6CajEPArEIigmuuuYZp06Ztc96VK1dWoKKMQ8CsgpYuXVq2/f3t5W0CZjnnkYBZGa1cuZK77rqLTz75BOCz3X61xCFgVkYffPABEyZM4MMPa/fkWl4dMMs5h4BZzjkEzHLOIWCWcw4Bs5xzCJjlnEPALOccAmY55xAwyzmHgFnOOQTMcs7fHTArg4hg/fr1NDY2lu16AR3FIWBWBps2beKSSy7h+eefZ82aNdUuZ6scAmZlEBEsXLiw5k4gUoy3CZjlXEkhIOlySa9Lek3SvZK6Sxoqaaakekn3p0uUmVmNancISBoIXAIMj4jDgDrgbOAm4JaIOAD4ELiwIwo1s/IodXWgC9BDUhdgF2AZcCLZZcoBJgNnlNiHmZVRKZcmXwr8CFhE9ub/CHgJWBURzSdWXwIMLPZ4SWMlzZY0u6Ghob1lmFmJSlkd2BM4HRgK7A3sCpzS1sdHxMSIGB4Rw/v27dveMsysRKWsDnwZeDsiGiJiI/AQcAywR1o9ABgELC2xRjMro1JCYBFwtKRdJAkYAcwFngHOTPOMAR4trUQzK6dStgnMJNsA+EdgTnquicB3gSsk1QN7AZM6oE4zK5OSjhiMiGuAa1o1vwUcWcrzmlnl+IhBs5xzCJjlnEPALOccAmY55xAwyzmHgFnOOQTMcs4hYJZzDgGznHMImOWcQ8As5xwCZjnnEDDLOYeAWc45BMxyziFglnMOAbOccwiY5ZxDwCznHAJmOecQMMs5h4BZzjkEzHLOIWCWcw4Bs5zbZghI+pmk5ZJeK2jrLelJSQvS7z1TuyTdJqle0quSjihn8WZWuraMBO7m85ccvwp4OiIOBJ5O9wFGAgemn7HAHR1TppmVyzZDICKeA1a2aj4dmJxuTwbOKGj/RWRmkF2mfEBHFWtmHa+92wT6RcSydPs9oF+6PRBYXDDfktT2OZLGSpotaXZDQ0M7yzCzUpW8YTAiAoh2PG5iRAyPiOF9+/YttQwza6f2hsD7zcP89Ht5al8KDC6Yb1Bq26pVq1bx4osvkuWJmVVSe0NgKjAm3R4DPFrQ/s20l+Bo4KOC1YYtWrhwITfddFM7SzGzUnTZ1gyS7gVOAPpIWgJcA0wApki6EHgXOCvN/htgFFAPfAyc39ZCPAowq45thkBEnLOFSSOKzBvAxaUWZWaV4yMGzXLOIWCWcw4Bs5zb5jaBSpk/fz433nhji7ZBgwZx3nnnUVdXV6WqzHZ8qoWt8pKKFnHMMccwffp0unbtWumSzErS1NTEySefzPTp06tdSqGXImJ460avDpjlnEPALOccAmY55xAwy7lOEQIR4cOKzcqkZnYRFvPmm28yZswYdtppJ+rq6hg/fjyHHHJItcsy26HUdAg0NDRw3333AdClSxcuuOACh4BZB+sUqwNmVj4OAbOccwiY5VxNbxMo9Omnn3LHHXcwbdq0bc572mmncdxxx1WgKrPOr9OEQEQwZcqUNs07aNAgh4BZG3l1wCznHAJmOecQMMs5h4BZznWaDYPbIyLYtGkTkpBU7XKshpXreymd6fsuNX1mofY6+OCDOeigg/jxj3/M/vvv35FPbTuYdevWcfnll7N48eJtz7ydXnzxRVaubH0t36oqemahHXIk8MYbb/Duu++yZs2aapdiNa6pqYnnnnuO+fPnV7uUqvE2AbOca8tlyH4GjAaWR8Rhqe1m4KvABmAhcH5ErErTxgMXAp8Cl0TE78pUu1m7LFq0iD/84Q8AfPzxx6xevbrKFVVXW1YH7gZuB35R0PYkMD4imiTdBIwHvitpGHA2cCiwN/CUpD+PiE87tmyz9ps5cybnnntutcuoGdtcHYiI54CVrdp+HxFN6e4MskuQA5wO3BcR6yPibbILkx7ZgfWaWQfriG0CFwC/TbcHAoWbWZekts+RNFbSbEmzO6AGM2unkvYOSLoaaALu2d7HRsREYGJ6nurvpzTLqXaHgKRvkW0wHBGbDzZYCgwumG1QajOzGtWu1QFJpwBXAqdFxMcFk6YCZ0vqJmkocCDwYullmlm5tGUX4b3ACUAfSUuAa8j2BnQDnkyH5c6IiL+LiNclTQHmkq0mXOw9A1ZNL7zwAo8//niLtjfeeKNK1dSmbYZARJxTpHnSVua/AbihlKLMOsqsWbP44Q9/WO0yapqPGDTLuR3yuwOQfYurvr6euro6APr27Uv//v2rXJVVSmNjI2+99RbLli2rdik1b4f8FmGzHj16sNNO2WDn8ssv5/rrry9HN1aD5syZw4gRI/joo4/YsGFDtcupFfn5FmGzxsbGz277hZAvmzZtYt26df6/t8EOHQKFip3kwScc2TEUG83Wwgi3s8hNCDz88MOffWe8f//+3HzzzfTs2bPKVVlHeOWVV7j22mvZtGnTZ22rV69m/fr1Vayq88hNCNTX11NfXw/Afvvtx8aNG6tckXWUhoYGpk6d2iIErO1qZcNgA7AOWFHtWoA+uI5CrqOlzlzHvhHRt3VjTYQAgKTZxbZcug7X4TrKW4cPFjLLOYeAWc7VUghMrHYBietoyXW0tMPVUTPbBMysOmppJGBmVeAQMMu5mggBSadImi+pXtJVFepzsKRnJM2V9LqkS1N7b0lPSlqQfu9ZoXrqJP1J0rR0f6ikmWmZ3C+pawVq2EPSA5LekDRP0peqsTwkXZ7+J69JuldS90otD0k/k7Rc0msFbUWXgTK3pZpelXREmeu4Of1vXpX0sKQ9CqaNT3XMl3TydnXWfEx9tX6AOrILmOwHdAVeAYZVoN8BwBHp9u7Am8Aw4J+Aq1L7VcBNFVoOVwC/Aqal+1OAs9PtO4GLKlDDZOBv0u2uwB6VXh5kZ6d+G+hRsBy+VanlARwHHAG8VtBWdBkAo8jOtC3gaGBmmes4CeiSbt9UUMew9L7pBgxN76e6NvdV7hdWG/7YLwG/K7g/nuzCJpWu41HgK8B8YEBqGwDMr0Dfg4CngROBaelFtaLgH95iGZWphl7pzadW7RVdHmw+bX1vssPapwEnV3J5AENavfmKLgPgX4Fzis1XjjpaTfsr4J50u8V7Bvgd8KW29lMLqwNtvlZBuUgaAhwOzAT6RUTzmSjeA/pVoISfkJ24tfng972AVbH5Ai+VWCZDgQbg52m15C5Ju1Lh5RERS4EfAYuAZcBHwEtUfnkU2tIyqOZrt13X+yimFkKgqiTtBjwIXBYRLS5KF1mslnUfqqTm6zy+VM5+2qAL2fDzjog4nOy7HC22z1RoeexJdiWroWSXstsVOKWcfW6PSiyDbSnleh/F1EIIVO1aBZJ2JguAeyLiodT8vqQBafoAYHmZyzgGOE3SO8B9ZKsEtwJ7SGr+lmcllskSYElEzEz3HyALhUovjy8Db0dEQ0RsBB4iW0aVXh6FtrQMKv7aLbjex3kpkEquoxZCYBZwYNr625XsgqZTy92psjOKTALmRcQ/F0yaCoxJt8eQbSsom4gYHxGDImII2d8+PSLOA54BzqxgHe8BiyUdlJpGkJ06vqLLg2w14GhJu6T/UXMdFV0erWxpGUwFvpn2EhwNfFSw2tDhyna9j3Ju5NmODSCjyLbOLwSurlCfx5IN614FXk4/o8jWx58GFgBPAb0ruBxOYPPegf3SP7Ie+DXQrQL9fxGYnZbJI8Ce1VgewLXAG8BrwC/JtnpXZHkA95Jti9hINjq6cEvLgGwD7k/T63YOMLzMddSTrfs3v17vLJj/6lTHfGDk9vTlw4bNcq4WVgfMrIocAmY55xAwyzmHgFnOOQTMcs4hYJZzDgGznPv/pkDpFnTuAW8AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "Q1gTC_IpH6yP",
        "outputId": "e5aa9f40-7767-4a30-f7cc-ca5ac38704b6"
      },
      "source": [
        "def display_examples(class_names, images, labels):\n",
        "    fig = plt.figure(figsize = (10,10))\n",
        "    fig.suptitle(\"Examples of images in the dataset\", fontsize=16)\n",
        "    for i in range(25):\n",
        "        plt.subplot(5,5,i+1)\n",
        "        plt.xticks([])\n",
        "        plt.yticks([])\n",
        "        plt.grid(False)\n",
        "        plt.imshow(images[i], cmap=plt.cm.binary)\n",
        "        plt.xlabel(class_names[labels[i]])\n",
        "    plt.show()\n",
        "    \n",
        "display_examples(class_names, train_images, train_labels)"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 25 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gkxGZHqQjJHZ"
      },
      "source": [
        "# EfficientNetV2\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F_W3s1NzCTSd"
      },
      "source": [
        "#Normalization\n",
        "train_images = train_images / 255.0 \n",
        "val_images = val_images / 255.0\n",
        "test_images = test_images / 255.0"
      ],
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WKvORLxgjINH"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import backend as K\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.layers import (\n",
        "    Activation,\n",
        "    Add,\n",
        "    BatchNormalization,\n",
        "    Conv2D,\n",
        "    Dense,\n",
        "    DepthwiseConv2D,\n",
        "    Dropout,\n",
        "    GlobalAveragePooling2D,\n",
        "    Input,\n",
        "    PReLU,\n",
        "    Reshape,\n",
        "    Multiply,\n",
        ")"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qHaE3aHijIL-"
      },
      "source": [
        "BATCH_NORM_DECAY = 0.9\n",
        "BATCH_NORM_EPSILON = 0.001\n",
        "CONV_KERNEL_INITIALIZER = keras.initializers.VarianceScaling(scale=2.0, mode=\"fan_out\", distribution=\"truncated_normal\")\n",
        "# CONV_KERNEL_INITIALIZER = 'glorot_uniform'\n",
        "\n",
        "BLOCK_CONFIGS = {\n",
        "    \"b0\": {\n",
        "        \"first_conv_filter\": 32,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6],\n",
        "        \"out_channels\": [16, 32, 48, 96, 112, 192],\n",
        "        \"depthes\": [1, 2, 2, 3, 5, 8],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1],\n",
        "    },\n",
        "    \"b1\": {\n",
        "        \"first_conv_filter\": 32,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6],\n",
        "        \"out_channels\": [16, 32, 48, 96, 112, 192],\n",
        "        \"depthes\": [2, 3, 3, 4, 6, 9],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1],\n",
        "    },\n",
        "    \"b2\": {\n",
        "        \"first_conv_filter\": 32,\n",
        "        \"output_conv_filter\": 1408,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6],\n",
        "        \"out_channels\": [16, 32, 56, 104, 120, 208],\n",
        "        \"depthes\": [2, 3, 3, 4, 6, 10],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1],\n",
        "    },\n",
        "    \"b3\": {\n",
        "        \"first_conv_filter\": 40,\n",
        "        \"output_conv_filter\": 1536,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6],\n",
        "        \"out_channels\": [16, 40, 56, 112, 136, 232],\n",
        "        \"depthes\": [2, 3, 3, 5, 7, 12],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1],\n",
        "    },\n",
        "    \"s\": {\n",
        "        \"first_conv_filter\": 24,\n",
        "        \"output_conv_filter\": 1280,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6],\n",
        "        \"out_channels\": [24, 48, 64, 128, 160, 256],\n",
        "        \"depthes\": [2, 4, 4, 6, 9, 15],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1],\n",
        "    },\n",
        "    \"m\": {\n",
        "        \"first_conv_filter\": 24,\n",
        "        \"output_conv_filter\": 1280,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6, 6],\n",
        "        \"out_channels\": [24, 48, 80, 160, 176, 304, 512],\n",
        "        \"depthes\": [3, 5, 5, 7, 14, 18, 5],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2, 1],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1, 1],\n",
        "    },\n",
        "    \"l\": {\n",
        "        \"first_conv_filter\": 32,\n",
        "        \"output_conv_filter\": 1280,\n",
        "        \"expands\": [1, 4, 4, 4, 6, 6, 6],\n",
        "        \"out_channels\": [32, 64, 96, 192, 224, 384, 640],\n",
        "        \"depthes\": [4, 7, 7, 10, 19, 25, 7],\n",
        "        \"strides\": [1, 2, 2, 2, 1, 2, 1],\n",
        "        \"use_ses\": [0, 0, 0, 1, 1, 1, 1],\n",
        "    },\n",
        "}"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "adznnDGLnFGA"
      },
      "source": [
        "def _make_divisible(v, divisor=4, min_value=None):\n",
        "    \"\"\"\n",
        "    This function is taken from the original tf repo.\n",
        "    It ensures that all layers have a channel number that is divisible by 8\n",
        "    It can be seen here:\n",
        "    https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py\n",
        "    \"\"\"\n",
        "    if min_value is None:\n",
        "        min_value = divisor\n",
        "    new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)\n",
        "    # Make sure that round down does not go down by more than 10%.\n",
        "    if new_v < 0.9 * v:\n",
        "        new_v += divisor\n",
        "    return new_v\n",
        "\n",
        "\n",
        "def conv2d_no_bias(inputs, filters, kernel_size, strides=1, padding=\"VALID\"):\n",
        "    return Conv2D(\n",
        "        filters, kernel_size, strides=strides, padding=padding, use_bias=False, kernel_initializer=CONV_KERNEL_INITIALIZER\n",
        "    )(inputs)\n",
        "\n",
        "\n",
        "def batchnorm_with_activation(inputs, activation=\"swish\"):\n",
        "    \"\"\"Performs a batch normalization followed by an activation. \"\"\"\n",
        "    bn_axis = 1 if K.image_data_format() == \"channels_first\" else -1\n",
        "    nn = BatchNormalization(\n",
        "        axis=bn_axis,\n",
        "        momentum=BATCH_NORM_DECAY,\n",
        "        epsilon=BATCH_NORM_EPSILON,\n",
        "    )(inputs)\n",
        "    if activation:\n",
        "        nn = Activation(activation=activation)(nn)\n",
        "        # nn = PReLU(shared_axes=[1, 2], alpha_initializer=tf.initializers.Constant(0.25))(nn)\n",
        "    return nn\n",
        "\n",
        "\n",
        "def se_module(inputs, se_ratio=4):\n",
        "    channel_axis = 1 if K.image_data_format() == \"channels_first\" else -1\n",
        "    h_axis, w_axis = [2, 3] if K.image_data_format() == \"channels_first\" else [1, 2]\n",
        "\n",
        "    filters = inputs.shape[channel_axis]\n",
        "    # reduction = _make_divisible(filters // se_ratio, 8)\n",
        "    reduction = filters // se_ratio\n",
        "    # se = GlobalAveragePooling2D()(inputs)\n",
        "    # se = Reshape((1, 1, filters))(se)\n",
        "    se = tf.reduce_mean(inputs, [h_axis, w_axis], keepdims=True)\n",
        "    se = Conv2D(reduction, kernel_size=1, use_bias=True, kernel_initializer=CONV_KERNEL_INITIALIZER)(se)\n",
        "    # se = PReLU(shared_axes=[1, 2])(se)\n",
        "    se = Activation(\"swish\")(se)\n",
        "    se = Conv2D(filters, kernel_size=1, use_bias=True, kernel_initializer=CONV_KERNEL_INITIALIZER)(se)\n",
        "    se = Activation(\"sigmoid\")(se)\n",
        "    return Multiply()([inputs, se])\n",
        "\n",
        "\n",
        "def MBConv(inputs, output_channel, stride, expand_ratio, shortcut, survival=None, use_se=0, is_fused=False):\n",
        "    channel_axis = 1 if K.image_data_format() == \"channels_first\" else -1\n",
        "    input_channel = inputs.shape[channel_axis]\n",
        "\n",
        "    if is_fused and expand_ratio != 1:\n",
        "        nn = conv2d_no_bias(inputs, input_channel * expand_ratio, (3, 3), strides=stride, padding=\"same\")\n",
        "        nn = batchnorm_with_activation(nn)\n",
        "    elif expand_ratio != 1:\n",
        "        nn = conv2d_no_bias(inputs, input_channel * expand_ratio, (1, 1), strides=(1, 1), padding=\"same\")\n",
        "        nn = batchnorm_with_activation(nn)\n",
        "    else:\n",
        "        nn = inputs\n",
        "\n",
        "    if not is_fused:\n",
        "        nn = DepthwiseConv2D(\n",
        "            (3, 3), padding=\"same\", strides=stride, use_bias=False, depthwise_initializer=CONV_KERNEL_INITIALIZER\n",
        "        )(nn)\n",
        "        nn = batchnorm_with_activation(nn)\n",
        "\n",
        "    if use_se:\n",
        "        nn = se_module(nn, se_ratio=4 * expand_ratio)\n",
        "\n",
        "    # pw-linear\n",
        "    if is_fused and expand_ratio == 1:\n",
        "        nn = conv2d_no_bias(nn, output_channel, (3, 3), strides=stride, padding=\"same\")\n",
        "        nn = batchnorm_with_activation(nn)\n",
        "    else:\n",
        "        nn = conv2d_no_bias(nn, output_channel, (1, 1), strides=(1, 1), padding=\"same\")\n",
        "        nn = batchnorm_with_activation(nn, activation=None)\n",
        "\n",
        "    if shortcut:\n",
        "        if survival is not None and survival < 1:\n",
        "            from tensorflow_addons.layers import StochasticDepth\n",
        "\n",
        "            return StochasticDepth(float(survival))([inputs, nn])\n",
        "        else:\n",
        "            return Add()([inputs, nn])\n",
        "    else:\n",
        "        return nn"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QzNt4EKthwJA"
      },
      "source": [
        "def EfficientNetV2(\n",
        "    model_type,\n",
        "    input_shape,\n",
        "    classes,\n",
        "    dropout=0.2,\n",
        "    first_strides=2,\n",
        "    survivals=None,\n",
        "    classifier_activation=\"softmax\",\n",
        "    name=\"EfficientNetV2\",\n",
        "    ):\n",
        "    \"\"\"\n",
        "    model_type: is the pre-defined model, value in [\"s\", \"m\", \"l\", \"b0\", \"b1\", \"b2\", \"b3\"].\n",
        "    first_strides: is used in the first Conv2D layer.\n",
        "    survivals: is used for [Deep Networks with Stochastic Depth](https://arxiv.org/abs/1603.09382).\n",
        "        Can be a constant value like `0.5` or `0.8`,\n",
        "        or a tuple value like `(1, 0.8)` indicates the survival probability linearly changes from `1 --> 0.8` for `top --> bottom` layers.\n",
        "        A higher value means a higher probability will keep the conv branch.\n",
        "        or `None` to disable.\n",
        "    \"\"\"\n",
        "    blocks_config = BLOCK_CONFIGS.get(model_type.lower(), BLOCK_CONFIGS[\"s\"])\n",
        "    expands = blocks_config[\"expands\"]\n",
        "    out_channels = blocks_config[\"out_channels\"]\n",
        "    depthes = blocks_config[\"depthes\"]\n",
        "    strides = blocks_config[\"strides\"]\n",
        "    use_ses = blocks_config[\"use_ses\"]\n",
        "    first_conv_filter = blocks_config.get(\"first_conv_filter\", out_channels[0])\n",
        "    output_conv_filter = blocks_config.get(\"output_conv_filter\", 1280)\n",
        "\n",
        "    inputs = Input(shape=input_shape)\n",
        "    out_channel = _make_divisible(first_conv_filter, 8)\n",
        "    nn = conv2d_no_bias(inputs, out_channel, (3, 3), strides=first_strides, padding=\"same\")\n",
        "    nn = batchnorm_with_activation(nn)\n",
        "\n",
        "    # StochasticDepth survival_probability values\n",
        "    total_layers = sum(depthes)\n",
        "    if isinstance(survivals, float):\n",
        "        survivals = [survivals] * total_layers\n",
        "    elif isinstance(survivals, (list, tuple)) and len(survivals) == 2:\n",
        "        start, end = survivals\n",
        "        survivals = [start - (1 - end) * float(ii) / total_layers for ii in range(total_layers)]\n",
        "    else:\n",
        "        survivals = [None] * total_layers\n",
        "    survivals = [survivals[int(sum(depthes[:id])) : sum(depthes[: id + 1])] for id in range(len(depthes))]\n",
        "\n",
        "    pre_out = out_channel\n",
        "    for expand, out_channel, depth, survival, stride, se in zip(expands, out_channels, depthes, survivals, strides, use_ses):\n",
        "        out = _make_divisible(out_channel, 8)\n",
        "        is_fused = True if se == 0 else False\n",
        "        for ii in range(depth):\n",
        "            stride = stride if ii == 0 else 1\n",
        "            shortcut = True if out == pre_out and stride == 1 else False\n",
        "            nn = MBConv(nn, out, stride, expand, shortcut, survival[ii], se, is_fused)\n",
        "            pre_out = out\n",
        "\n",
        "    output_conv_filter = _make_divisible(output_conv_filter, 8)\n",
        "    nn = conv2d_no_bias(nn, output_conv_filter, (1, 1), strides=(1, 1), padding=\"valid\")\n",
        "    nn = batchnorm_with_activation(nn)\n",
        "\n",
        "    if classes > 0:\n",
        "        nn = GlobalAveragePooling2D(name=\"avg_pool\")(nn)\n",
        "        if dropout > 0 and dropout < 1:\n",
        "            nn = Dropout(dropout)(nn)\n",
        "        nn = Dense(classes, activation=classifier_activation, name=\"predictions\")(nn)\n",
        "    return Model(inputs=inputs, outputs=nn, name=name)"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GsX57sYQpyi-",
        "outputId": "8803df70-a704-4be5-d478-6fa41ffa96ff"
      },
      "source": [
        "model = EfficientNetV2(\n",
        "    model_type=\"b3\",\n",
        "    input_shape=(IMAGE_SIZE, IMAGE_SIZE, 3),\n",
        "    classes=4,\n",
        "    dropout=0.2,\n",
        "    first_strides=2,\n",
        "    survivals=None,\n",
        "    classifier_activation=\"softmax\",\n",
        "    name=\"EfficientNetV2\",\n",
        ")\n",
        "model.summary()"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"EfficientNetV2\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 128, 128, 3) 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d (Conv2D)                 (None, 64, 64, 40)   1080        input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization (BatchNorma (None, 64, 64, 40)   160         conv2d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "activation (Activation)         (None, 64, 64, 40)   0           batch_normalization[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 64, 64, 16)   5760        activation[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_1 (BatchNor (None, 64, 64, 16)   64          conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_1 (Activation)       (None, 64, 64, 16)   0           batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 64, 64, 16)   2304        activation_1[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_2 (BatchNor (None, 64, 64, 16)   64          conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_2 (Activation)       (None, 64, 64, 16)   0           batch_normalization_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "add (Add)                       (None, 64, 64, 16)   0           activation_1[0][0]               \n",
            "                                                                 activation_2[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 32, 32, 64)   9216        add[0][0]                        \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_3 (BatchNor (None, 32, 32, 64)   256         conv2d_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_3 (Activation)       (None, 32, 32, 64)   0           batch_normalization_3[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_4 (Conv2D)               (None, 32, 32, 40)   2560        activation_3[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_4 (BatchNor (None, 32, 32, 40)   160         conv2d_4[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_5 (Conv2D)               (None, 32, 32, 160)  57600       batch_normalization_4[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_5 (BatchNor (None, 32, 32, 160)  640         conv2d_5[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_4 (Activation)       (None, 32, 32, 160)  0           batch_normalization_5[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_6 (Conv2D)               (None, 32, 32, 40)   6400        activation_4[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_6 (BatchNor (None, 32, 32, 40)   160         conv2d_6[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "add_1 (Add)                     (None, 32, 32, 40)   0           batch_normalization_4[0][0]      \n",
            "                                                                 batch_normalization_6[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_7 (Conv2D)               (None, 32, 32, 160)  57600       add_1[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_7 (BatchNor (None, 32, 32, 160)  640         conv2d_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_5 (Activation)       (None, 32, 32, 160)  0           batch_normalization_7[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_8 (Conv2D)               (None, 32, 32, 40)   6400        activation_5[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_8 (BatchNor (None, 32, 32, 40)   160         conv2d_8[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "add_2 (Add)                     (None, 32, 32, 40)   0           add_1[0][0]                      \n",
            "                                                                 batch_normalization_8[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_9 (Conv2D)               (None, 16, 16, 160)  57600       add_2[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_9 (BatchNor (None, 16, 16, 160)  640         conv2d_9[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_6 (Activation)       (None, 16, 16, 160)  0           batch_normalization_9[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_10 (Conv2D)              (None, 16, 16, 56)   8960        activation_6[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_10 (BatchNo (None, 16, 16, 56)   224         conv2d_10[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_11 (Conv2D)              (None, 16, 16, 224)  112896      batch_normalization_10[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_11 (BatchNo (None, 16, 16, 224)  896         conv2d_11[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_7 (Activation)       (None, 16, 16, 224)  0           batch_normalization_11[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_12 (Conv2D)              (None, 16, 16, 56)   12544       activation_7[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_12 (BatchNo (None, 16, 16, 56)   224         conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_3 (Add)                     (None, 16, 16, 56)   0           batch_normalization_10[0][0]     \n",
            "                                                                 batch_normalization_12[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_13 (Conv2D)              (None, 16, 16, 224)  112896      add_3[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_13 (BatchNo (None, 16, 16, 224)  896         conv2d_13[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_8 (Activation)       (None, 16, 16, 224)  0           batch_normalization_13[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_14 (Conv2D)              (None, 16, 16, 56)   12544       activation_8[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_14 (BatchNo (None, 16, 16, 56)   224         conv2d_14[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_4 (Add)                     (None, 16, 16, 56)   0           add_3[0][0]                      \n",
            "                                                                 batch_normalization_14[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_15 (Conv2D)              (None, 16, 16, 224)  12544       add_4[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_15 (BatchNo (None, 16, 16, 224)  896         conv2d_15[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_9 (Activation)       (None, 16, 16, 224)  0           batch_normalization_15[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d (DepthwiseConv (None, 8, 8, 224)    2016        activation_9[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_16 (BatchNo (None, 8, 8, 224)    896         depthwise_conv2d[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_10 (Activation)      (None, 8, 8, 224)    0           batch_normalization_16[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean (TFOpLambda (None, 1, 1, 224)    0           activation_10[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_16 (Conv2D)              (None, 1, 1, 14)     3150        tf.math.reduce_mean[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_11 (Activation)      (None, 1, 1, 14)     0           conv2d_16[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_17 (Conv2D)              (None, 1, 1, 224)    3360        activation_11[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_12 (Activation)      (None, 1, 1, 224)    0           conv2d_17[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply (Multiply)             (None, 8, 8, 224)    0           activation_10[0][0]              \n",
            "                                                                 activation_12[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_18 (Conv2D)              (None, 8, 8, 112)    25088       multiply[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_17 (BatchNo (None, 8, 8, 112)    448         conv2d_18[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_19 (Conv2D)              (None, 8, 8, 448)    50176       batch_normalization_17[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_18 (BatchNo (None, 8, 8, 448)    1792        conv2d_19[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_13 (Activation)      (None, 8, 8, 448)    0           batch_normalization_18[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_1 (DepthwiseCo (None, 8, 8, 448)    4032        activation_13[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_19 (BatchNo (None, 8, 8, 448)    1792        depthwise_conv2d_1[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_14 (Activation)      (None, 8, 8, 448)    0           batch_normalization_19[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_1 (TFOpLamb (None, 1, 1, 448)    0           activation_14[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_20 (Conv2D)              (None, 1, 1, 28)     12572       tf.math.reduce_mean_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_15 (Activation)      (None, 1, 1, 28)     0           conv2d_20[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_21 (Conv2D)              (None, 1, 1, 448)    12992       activation_15[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_16 (Activation)      (None, 1, 1, 448)    0           conv2d_21[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_1 (Multiply)           (None, 8, 8, 448)    0           activation_14[0][0]              \n",
            "                                                                 activation_16[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_22 (Conv2D)              (None, 8, 8, 112)    50176       multiply_1[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_20 (BatchNo (None, 8, 8, 112)    448         conv2d_22[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_5 (Add)                     (None, 8, 8, 112)    0           batch_normalization_17[0][0]     \n",
            "                                                                 batch_normalization_20[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_23 (Conv2D)              (None, 8, 8, 448)    50176       add_5[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_21 (BatchNo (None, 8, 8, 448)    1792        conv2d_23[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_17 (Activation)      (None, 8, 8, 448)    0           batch_normalization_21[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_2 (DepthwiseCo (None, 8, 8, 448)    4032        activation_17[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_22 (BatchNo (None, 8, 8, 448)    1792        depthwise_conv2d_2[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_18 (Activation)      (None, 8, 8, 448)    0           batch_normalization_22[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_2 (TFOpLamb (None, 1, 1, 448)    0           activation_18[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_24 (Conv2D)              (None, 1, 1, 28)     12572       tf.math.reduce_mean_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_19 (Activation)      (None, 1, 1, 28)     0           conv2d_24[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_25 (Conv2D)              (None, 1, 1, 448)    12992       activation_19[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_20 (Activation)      (None, 1, 1, 448)    0           conv2d_25[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_2 (Multiply)           (None, 8, 8, 448)    0           activation_18[0][0]              \n",
            "                                                                 activation_20[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_26 (Conv2D)              (None, 8, 8, 112)    50176       multiply_2[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_23 (BatchNo (None, 8, 8, 112)    448         conv2d_26[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_6 (Add)                     (None, 8, 8, 112)    0           add_5[0][0]                      \n",
            "                                                                 batch_normalization_23[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_27 (Conv2D)              (None, 8, 8, 448)    50176       add_6[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_24 (BatchNo (None, 8, 8, 448)    1792        conv2d_27[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_21 (Activation)      (None, 8, 8, 448)    0           batch_normalization_24[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_3 (DepthwiseCo (None, 8, 8, 448)    4032        activation_21[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_25 (BatchNo (None, 8, 8, 448)    1792        depthwise_conv2d_3[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_22 (Activation)      (None, 8, 8, 448)    0           batch_normalization_25[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_3 (TFOpLamb (None, 1, 1, 448)    0           activation_22[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_28 (Conv2D)              (None, 1, 1, 28)     12572       tf.math.reduce_mean_3[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_23 (Activation)      (None, 1, 1, 28)     0           conv2d_28[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_29 (Conv2D)              (None, 1, 1, 448)    12992       activation_23[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_24 (Activation)      (None, 1, 1, 448)    0           conv2d_29[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_3 (Multiply)           (None, 8, 8, 448)    0           activation_22[0][0]              \n",
            "                                                                 activation_24[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_30 (Conv2D)              (None, 8, 8, 112)    50176       multiply_3[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_26 (BatchNo (None, 8, 8, 112)    448         conv2d_30[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_7 (Add)                     (None, 8, 8, 112)    0           add_6[0][0]                      \n",
            "                                                                 batch_normalization_26[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_31 (Conv2D)              (None, 8, 8, 448)    50176       add_7[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_27 (BatchNo (None, 8, 8, 448)    1792        conv2d_31[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_25 (Activation)      (None, 8, 8, 448)    0           batch_normalization_27[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_4 (DepthwiseCo (None, 8, 8, 448)    4032        activation_25[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_28 (BatchNo (None, 8, 8, 448)    1792        depthwise_conv2d_4[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_26 (Activation)      (None, 8, 8, 448)    0           batch_normalization_28[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_4 (TFOpLamb (None, 1, 1, 448)    0           activation_26[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_32 (Conv2D)              (None, 1, 1, 28)     12572       tf.math.reduce_mean_4[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_27 (Activation)      (None, 1, 1, 28)     0           conv2d_32[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_33 (Conv2D)              (None, 1, 1, 448)    12992       activation_27[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_28 (Activation)      (None, 1, 1, 448)    0           conv2d_33[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_4 (Multiply)           (None, 8, 8, 448)    0           activation_26[0][0]              \n",
            "                                                                 activation_28[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_34 (Conv2D)              (None, 8, 8, 112)    50176       multiply_4[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_29 (BatchNo (None, 8, 8, 112)    448         conv2d_34[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_8 (Add)                     (None, 8, 8, 112)    0           add_7[0][0]                      \n",
            "                                                                 batch_normalization_29[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_35 (Conv2D)              (None, 8, 8, 672)    75264       add_8[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_30 (BatchNo (None, 8, 8, 672)    2688        conv2d_35[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_29 (Activation)      (None, 8, 8, 672)    0           batch_normalization_30[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_5 (DepthwiseCo (None, 8, 8, 672)    6048        activation_29[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_31 (BatchNo (None, 8, 8, 672)    2688        depthwise_conv2d_5[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_30 (Activation)      (None, 8, 8, 672)    0           batch_normalization_31[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_5 (TFOpLamb (None, 1, 1, 672)    0           activation_30[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_36 (Conv2D)              (None, 1, 1, 28)     18844       tf.math.reduce_mean_5[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_31 (Activation)      (None, 1, 1, 28)     0           conv2d_36[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_37 (Conv2D)              (None, 1, 1, 672)    19488       activation_31[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_32 (Activation)      (None, 1, 1, 672)    0           conv2d_37[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_5 (Multiply)           (None, 8, 8, 672)    0           activation_30[0][0]              \n",
            "                                                                 activation_32[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_38 (Conv2D)              (None, 8, 8, 136)    91392       multiply_5[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_32 (BatchNo (None, 8, 8, 136)    544         conv2d_38[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_39 (Conv2D)              (None, 8, 8, 816)    110976      batch_normalization_32[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_33 (BatchNo (None, 8, 8, 816)    3264        conv2d_39[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_33 (Activation)      (None, 8, 8, 816)    0           batch_normalization_33[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_6 (DepthwiseCo (None, 8, 8, 816)    7344        activation_33[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_34 (BatchNo (None, 8, 8, 816)    3264        depthwise_conv2d_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_34 (Activation)      (None, 8, 8, 816)    0           batch_normalization_34[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_6 (TFOpLamb (None, 1, 1, 816)    0           activation_34[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_40 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_6[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_35 (Activation)      (None, 1, 1, 34)     0           conv2d_40[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_41 (Conv2D)              (None, 1, 1, 816)    28560       activation_35[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_36 (Activation)      (None, 1, 1, 816)    0           conv2d_41[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_6 (Multiply)           (None, 8, 8, 816)    0           activation_34[0][0]              \n",
            "                                                                 activation_36[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_42 (Conv2D)              (None, 8, 8, 136)    110976      multiply_6[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_35 (BatchNo (None, 8, 8, 136)    544         conv2d_42[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_9 (Add)                     (None, 8, 8, 136)    0           batch_normalization_32[0][0]     \n",
            "                                                                 batch_normalization_35[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_43 (Conv2D)              (None, 8, 8, 816)    110976      add_9[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_36 (BatchNo (None, 8, 8, 816)    3264        conv2d_43[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_37 (Activation)      (None, 8, 8, 816)    0           batch_normalization_36[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_7 (DepthwiseCo (None, 8, 8, 816)    7344        activation_37[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_37 (BatchNo (None, 8, 8, 816)    3264        depthwise_conv2d_7[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_38 (Activation)      (None, 8, 8, 816)    0           batch_normalization_37[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_7 (TFOpLamb (None, 1, 1, 816)    0           activation_38[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_44 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_7[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_39 (Activation)      (None, 1, 1, 34)     0           conv2d_44[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_45 (Conv2D)              (None, 1, 1, 816)    28560       activation_39[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_40 (Activation)      (None, 1, 1, 816)    0           conv2d_45[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_7 (Multiply)           (None, 8, 8, 816)    0           activation_38[0][0]              \n",
            "                                                                 activation_40[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_46 (Conv2D)              (None, 8, 8, 136)    110976      multiply_7[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_38 (BatchNo (None, 8, 8, 136)    544         conv2d_46[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_10 (Add)                    (None, 8, 8, 136)    0           add_9[0][0]                      \n",
            "                                                                 batch_normalization_38[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_47 (Conv2D)              (None, 8, 8, 816)    110976      add_10[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_39 (BatchNo (None, 8, 8, 816)    3264        conv2d_47[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_41 (Activation)      (None, 8, 8, 816)    0           batch_normalization_39[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_8 (DepthwiseCo (None, 8, 8, 816)    7344        activation_41[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_40 (BatchNo (None, 8, 8, 816)    3264        depthwise_conv2d_8[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_42 (Activation)      (None, 8, 8, 816)    0           batch_normalization_40[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_8 (TFOpLamb (None, 1, 1, 816)    0           activation_42[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_48 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_8[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_43 (Activation)      (None, 1, 1, 34)     0           conv2d_48[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_49 (Conv2D)              (None, 1, 1, 816)    28560       activation_43[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_44 (Activation)      (None, 1, 1, 816)    0           conv2d_49[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_8 (Multiply)           (None, 8, 8, 816)    0           activation_42[0][0]              \n",
            "                                                                 activation_44[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_50 (Conv2D)              (None, 8, 8, 136)    110976      multiply_8[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_41 (BatchNo (None, 8, 8, 136)    544         conv2d_50[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_11 (Add)                    (None, 8, 8, 136)    0           add_10[0][0]                     \n",
            "                                                                 batch_normalization_41[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_51 (Conv2D)              (None, 8, 8, 816)    110976      add_11[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_42 (BatchNo (None, 8, 8, 816)    3264        conv2d_51[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_45 (Activation)      (None, 8, 8, 816)    0           batch_normalization_42[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_9 (DepthwiseCo (None, 8, 8, 816)    7344        activation_45[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_43 (BatchNo (None, 8, 8, 816)    3264        depthwise_conv2d_9[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "activation_46 (Activation)      (None, 8, 8, 816)    0           batch_normalization_43[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_9 (TFOpLamb (None, 1, 1, 816)    0           activation_46[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_52 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_9[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_47 (Activation)      (None, 1, 1, 34)     0           conv2d_52[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_53 (Conv2D)              (None, 1, 1, 816)    28560       activation_47[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_48 (Activation)      (None, 1, 1, 816)    0           conv2d_53[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_9 (Multiply)           (None, 8, 8, 816)    0           activation_46[0][0]              \n",
            "                                                                 activation_48[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_54 (Conv2D)              (None, 8, 8, 136)    110976      multiply_9[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_44 (BatchNo (None, 8, 8, 136)    544         conv2d_54[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_12 (Add)                    (None, 8, 8, 136)    0           add_11[0][0]                     \n",
            "                                                                 batch_normalization_44[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_55 (Conv2D)              (None, 8, 8, 816)    110976      add_12[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_45 (BatchNo (None, 8, 8, 816)    3264        conv2d_55[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_49 (Activation)      (None, 8, 8, 816)    0           batch_normalization_45[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_10 (DepthwiseC (None, 8, 8, 816)    7344        activation_49[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_46 (BatchNo (None, 8, 8, 816)    3264        depthwise_conv2d_10[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_50 (Activation)      (None, 8, 8, 816)    0           batch_normalization_46[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_10 (TFOpLam (None, 1, 1, 816)    0           activation_50[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_56 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_10[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_51 (Activation)      (None, 1, 1, 34)     0           conv2d_56[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_57 (Conv2D)              (None, 1, 1, 816)    28560       activation_51[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_52 (Activation)      (None, 1, 1, 816)    0           conv2d_57[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_10 (Multiply)          (None, 8, 8, 816)    0           activation_50[0][0]              \n",
            "                                                                 activation_52[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_58 (Conv2D)              (None, 8, 8, 136)    110976      multiply_10[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_47 (BatchNo (None, 8, 8, 136)    544         conv2d_58[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_13 (Add)                    (None, 8, 8, 136)    0           add_12[0][0]                     \n",
            "                                                                 batch_normalization_47[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_59 (Conv2D)              (None, 8, 8, 816)    110976      add_13[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_48 (BatchNo (None, 8, 8, 816)    3264        conv2d_59[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_53 (Activation)      (None, 8, 8, 816)    0           batch_normalization_48[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_11 (DepthwiseC (None, 8, 8, 816)    7344        activation_53[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_49 (BatchNo (None, 8, 8, 816)    3264        depthwise_conv2d_11[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_54 (Activation)      (None, 8, 8, 816)    0           batch_normalization_49[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_11 (TFOpLam (None, 1, 1, 816)    0           activation_54[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_60 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_11[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_55 (Activation)      (None, 1, 1, 34)     0           conv2d_60[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_61 (Conv2D)              (None, 1, 1, 816)    28560       activation_55[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_56 (Activation)      (None, 1, 1, 816)    0           conv2d_61[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_11 (Multiply)          (None, 8, 8, 816)    0           activation_54[0][0]              \n",
            "                                                                 activation_56[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_62 (Conv2D)              (None, 8, 8, 136)    110976      multiply_11[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_50 (BatchNo (None, 8, 8, 136)    544         conv2d_62[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_14 (Add)                    (None, 8, 8, 136)    0           add_13[0][0]                     \n",
            "                                                                 batch_normalization_50[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_63 (Conv2D)              (None, 8, 8, 816)    110976      add_14[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_51 (BatchNo (None, 8, 8, 816)    3264        conv2d_63[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_57 (Activation)      (None, 8, 8, 816)    0           batch_normalization_51[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_12 (DepthwiseC (None, 4, 4, 816)    7344        activation_57[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_52 (BatchNo (None, 4, 4, 816)    3264        depthwise_conv2d_12[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_58 (Activation)      (None, 4, 4, 816)    0           batch_normalization_52[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_12 (TFOpLam (None, 1, 1, 816)    0           activation_58[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_64 (Conv2D)              (None, 1, 1, 34)     27778       tf.math.reduce_mean_12[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_59 (Activation)      (None, 1, 1, 34)     0           conv2d_64[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_65 (Conv2D)              (None, 1, 1, 816)    28560       activation_59[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_60 (Activation)      (None, 1, 1, 816)    0           conv2d_65[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_12 (Multiply)          (None, 4, 4, 816)    0           activation_58[0][0]              \n",
            "                                                                 activation_60[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_66 (Conv2D)              (None, 4, 4, 232)    189312      multiply_12[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_53 (BatchNo (None, 4, 4, 232)    928         conv2d_66[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_67 (Conv2D)              (None, 4, 4, 1392)   322944      batch_normalization_53[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_54 (BatchNo (None, 4, 4, 1392)   5568        conv2d_67[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_61 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_54[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_13 (DepthwiseC (None, 4, 4, 1392)   12528       activation_61[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_55 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_13[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_62 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_55[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_13 (TFOpLam (None, 1, 1, 1392)   0           activation_62[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_68 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_13[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_63 (Activation)      (None, 1, 1, 58)     0           conv2d_68[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_69 (Conv2D)              (None, 1, 1, 1392)   82128       activation_63[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_64 (Activation)      (None, 1, 1, 1392)   0           conv2d_69[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_13 (Multiply)          (None, 4, 4, 1392)   0           activation_62[0][0]              \n",
            "                                                                 activation_64[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_70 (Conv2D)              (None, 4, 4, 232)    322944      multiply_13[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_56 (BatchNo (None, 4, 4, 232)    928         conv2d_70[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_15 (Add)                    (None, 4, 4, 232)    0           batch_normalization_53[0][0]     \n",
            "                                                                 batch_normalization_56[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_71 (Conv2D)              (None, 4, 4, 1392)   322944      add_15[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_57 (BatchNo (None, 4, 4, 1392)   5568        conv2d_71[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_65 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_57[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_14 (DepthwiseC (None, 4, 4, 1392)   12528       activation_65[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_58 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_14[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_66 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_58[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_14 (TFOpLam (None, 1, 1, 1392)   0           activation_66[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_72 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_14[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_67 (Activation)      (None, 1, 1, 58)     0           conv2d_72[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_73 (Conv2D)              (None, 1, 1, 1392)   82128       activation_67[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_68 (Activation)      (None, 1, 1, 1392)   0           conv2d_73[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_14 (Multiply)          (None, 4, 4, 1392)   0           activation_66[0][0]              \n",
            "                                                                 activation_68[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_74 (Conv2D)              (None, 4, 4, 232)    322944      multiply_14[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_59 (BatchNo (None, 4, 4, 232)    928         conv2d_74[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_16 (Add)                    (None, 4, 4, 232)    0           add_15[0][0]                     \n",
            "                                                                 batch_normalization_59[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_75 (Conv2D)              (None, 4, 4, 1392)   322944      add_16[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_60 (BatchNo (None, 4, 4, 1392)   5568        conv2d_75[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_69 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_60[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_15 (DepthwiseC (None, 4, 4, 1392)   12528       activation_69[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_61 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_15[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_70 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_61[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_15 (TFOpLam (None, 1, 1, 1392)   0           activation_70[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_76 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_15[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_71 (Activation)      (None, 1, 1, 58)     0           conv2d_76[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_77 (Conv2D)              (None, 1, 1, 1392)   82128       activation_71[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_72 (Activation)      (None, 1, 1, 1392)   0           conv2d_77[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_15 (Multiply)          (None, 4, 4, 1392)   0           activation_70[0][0]              \n",
            "                                                                 activation_72[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_78 (Conv2D)              (None, 4, 4, 232)    322944      multiply_15[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_62 (BatchNo (None, 4, 4, 232)    928         conv2d_78[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_17 (Add)                    (None, 4, 4, 232)    0           add_16[0][0]                     \n",
            "                                                                 batch_normalization_62[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_79 (Conv2D)              (None, 4, 4, 1392)   322944      add_17[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_63 (BatchNo (None, 4, 4, 1392)   5568        conv2d_79[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_73 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_63[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_16 (DepthwiseC (None, 4, 4, 1392)   12528       activation_73[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_64 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_16[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_74 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_64[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_16 (TFOpLam (None, 1, 1, 1392)   0           activation_74[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_80 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_16[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_75 (Activation)      (None, 1, 1, 58)     0           conv2d_80[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_81 (Conv2D)              (None, 1, 1, 1392)   82128       activation_75[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_76 (Activation)      (None, 1, 1, 1392)   0           conv2d_81[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_16 (Multiply)          (None, 4, 4, 1392)   0           activation_74[0][0]              \n",
            "                                                                 activation_76[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_82 (Conv2D)              (None, 4, 4, 232)    322944      multiply_16[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_65 (BatchNo (None, 4, 4, 232)    928         conv2d_82[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_18 (Add)                    (None, 4, 4, 232)    0           add_17[0][0]                     \n",
            "                                                                 batch_normalization_65[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_83 (Conv2D)              (None, 4, 4, 1392)   322944      add_18[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_66 (BatchNo (None, 4, 4, 1392)   5568        conv2d_83[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_77 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_66[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_17 (DepthwiseC (None, 4, 4, 1392)   12528       activation_77[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_67 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_17[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_78 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_67[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_17 (TFOpLam (None, 1, 1, 1392)   0           activation_78[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_84 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_17[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_79 (Activation)      (None, 1, 1, 58)     0           conv2d_84[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_85 (Conv2D)              (None, 1, 1, 1392)   82128       activation_79[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_80 (Activation)      (None, 1, 1, 1392)   0           conv2d_85[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_17 (Multiply)          (None, 4, 4, 1392)   0           activation_78[0][0]              \n",
            "                                                                 activation_80[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_86 (Conv2D)              (None, 4, 4, 232)    322944      multiply_17[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_68 (BatchNo (None, 4, 4, 232)    928         conv2d_86[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_19 (Add)                    (None, 4, 4, 232)    0           add_18[0][0]                     \n",
            "                                                                 batch_normalization_68[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_87 (Conv2D)              (None, 4, 4, 1392)   322944      add_19[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_69 (BatchNo (None, 4, 4, 1392)   5568        conv2d_87[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_81 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_69[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_18 (DepthwiseC (None, 4, 4, 1392)   12528       activation_81[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_70 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_18[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_82 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_70[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_18 (TFOpLam (None, 1, 1, 1392)   0           activation_82[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_88 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_18[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_83 (Activation)      (None, 1, 1, 58)     0           conv2d_88[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_89 (Conv2D)              (None, 1, 1, 1392)   82128       activation_83[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_84 (Activation)      (None, 1, 1, 1392)   0           conv2d_89[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_18 (Multiply)          (None, 4, 4, 1392)   0           activation_82[0][0]              \n",
            "                                                                 activation_84[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_90 (Conv2D)              (None, 4, 4, 232)    322944      multiply_18[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_71 (BatchNo (None, 4, 4, 232)    928         conv2d_90[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_20 (Add)                    (None, 4, 4, 232)    0           add_19[0][0]                     \n",
            "                                                                 batch_normalization_71[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_91 (Conv2D)              (None, 4, 4, 1392)   322944      add_20[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_72 (BatchNo (None, 4, 4, 1392)   5568        conv2d_91[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_85 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_72[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_19 (DepthwiseC (None, 4, 4, 1392)   12528       activation_85[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_73 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_19[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_86 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_73[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_19 (TFOpLam (None, 1, 1, 1392)   0           activation_86[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_92 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_19[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_87 (Activation)      (None, 1, 1, 58)     0           conv2d_92[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_93 (Conv2D)              (None, 1, 1, 1392)   82128       activation_87[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_88 (Activation)      (None, 1, 1, 1392)   0           conv2d_93[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_19 (Multiply)          (None, 4, 4, 1392)   0           activation_86[0][0]              \n",
            "                                                                 activation_88[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_94 (Conv2D)              (None, 4, 4, 232)    322944      multiply_19[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_74 (BatchNo (None, 4, 4, 232)    928         conv2d_94[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_21 (Add)                    (None, 4, 4, 232)    0           add_20[0][0]                     \n",
            "                                                                 batch_normalization_74[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_95 (Conv2D)              (None, 4, 4, 1392)   322944      add_21[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_75 (BatchNo (None, 4, 4, 1392)   5568        conv2d_95[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_89 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_75[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_20 (DepthwiseC (None, 4, 4, 1392)   12528       activation_89[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_76 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_20[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_90 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_76[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_20 (TFOpLam (None, 1, 1, 1392)   0           activation_90[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_96 (Conv2D)              (None, 1, 1, 58)     80794       tf.math.reduce_mean_20[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_91 (Activation)      (None, 1, 1, 58)     0           conv2d_96[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_97 (Conv2D)              (None, 1, 1, 1392)   82128       activation_91[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_92 (Activation)      (None, 1, 1, 1392)   0           conv2d_97[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "multiply_20 (Multiply)          (None, 4, 4, 1392)   0           activation_90[0][0]              \n",
            "                                                                 activation_92[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_98 (Conv2D)              (None, 4, 4, 232)    322944      multiply_20[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_77 (BatchNo (None, 4, 4, 232)    928         conv2d_98[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "add_22 (Add)                    (None, 4, 4, 232)    0           add_21[0][0]                     \n",
            "                                                                 batch_normalization_77[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_99 (Conv2D)              (None, 4, 4, 1392)   322944      add_22[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_78 (BatchNo (None, 4, 4, 1392)   5568        conv2d_99[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_93 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_78[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_21 (DepthwiseC (None, 4, 4, 1392)   12528       activation_93[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_79 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_21[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_94 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_79[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_21 (TFOpLam (None, 1, 1, 1392)   0           activation_94[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_100 (Conv2D)             (None, 1, 1, 58)     80794       tf.math.reduce_mean_21[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_95 (Activation)      (None, 1, 1, 58)     0           conv2d_100[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_101 (Conv2D)             (None, 1, 1, 1392)   82128       activation_95[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_96 (Activation)      (None, 1, 1, 1392)   0           conv2d_101[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "multiply_21 (Multiply)          (None, 4, 4, 1392)   0           activation_94[0][0]              \n",
            "                                                                 activation_96[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_102 (Conv2D)             (None, 4, 4, 232)    322944      multiply_21[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_80 (BatchNo (None, 4, 4, 232)    928         conv2d_102[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "add_23 (Add)                    (None, 4, 4, 232)    0           add_22[0][0]                     \n",
            "                                                                 batch_normalization_80[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_103 (Conv2D)             (None, 4, 4, 1392)   322944      add_23[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_81 (BatchNo (None, 4, 4, 1392)   5568        conv2d_103[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_97 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_81[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_22 (DepthwiseC (None, 4, 4, 1392)   12528       activation_97[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_82 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_22[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_98 (Activation)      (None, 4, 4, 1392)   0           batch_normalization_82[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_22 (TFOpLam (None, 1, 1, 1392)   0           activation_98[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_104 (Conv2D)             (None, 1, 1, 58)     80794       tf.math.reduce_mean_22[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_99 (Activation)      (None, 1, 1, 58)     0           conv2d_104[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_105 (Conv2D)             (None, 1, 1, 1392)   82128       activation_99[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_100 (Activation)     (None, 1, 1, 1392)   0           conv2d_105[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "multiply_22 (Multiply)          (None, 4, 4, 1392)   0           activation_98[0][0]              \n",
            "                                                                 activation_100[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_106 (Conv2D)             (None, 4, 4, 232)    322944      multiply_22[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_83 (BatchNo (None, 4, 4, 232)    928         conv2d_106[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "add_24 (Add)                    (None, 4, 4, 232)    0           add_23[0][0]                     \n",
            "                                                                 batch_normalization_83[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_107 (Conv2D)             (None, 4, 4, 1392)   322944      add_24[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_84 (BatchNo (None, 4, 4, 1392)   5568        conv2d_107[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_101 (Activation)     (None, 4, 4, 1392)   0           batch_normalization_84[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "depthwise_conv2d_23 (DepthwiseC (None, 4, 4, 1392)   12528       activation_101[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_85 (BatchNo (None, 4, 4, 1392)   5568        depthwise_conv2d_23[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_102 (Activation)     (None, 4, 4, 1392)   0           batch_normalization_85[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_23 (TFOpLam (None, 1, 1, 1392)   0           activation_102[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_108 (Conv2D)             (None, 1, 1, 58)     80794       tf.math.reduce_mean_23[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_103 (Activation)     (None, 1, 1, 58)     0           conv2d_108[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_109 (Conv2D)             (None, 1, 1, 1392)   82128       activation_103[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_104 (Activation)     (None, 1, 1, 1392)   0           conv2d_109[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "multiply_23 (Multiply)          (None, 4, 4, 1392)   0           activation_102[0][0]             \n",
            "                                                                 activation_104[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_110 (Conv2D)             (None, 4, 4, 232)    322944      multiply_23[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_86 (BatchNo (None, 4, 4, 232)    928         conv2d_110[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "add_25 (Add)                    (None, 4, 4, 232)    0           add_24[0][0]                     \n",
            "                                                                 batch_normalization_86[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_111 (Conv2D)             (None, 4, 4, 1536)   356352      add_25[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_87 (BatchNo (None, 4, 4, 1536)   6144        conv2d_111[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_105 (Activation)     (None, 4, 4, 1536)   0           batch_normalization_87[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "avg_pool (GlobalAveragePooling2 (None, 1536)         0           activation_105[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "dropout (Dropout)               (None, 1536)         0           avg_pool[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "predictions (Dense)             (None, 4)            6148        dropout[0][0]                    \n",
            "==================================================================================================\n",
            "Total params: 12,936,770\n",
            "Trainable params: 12,827,554\n",
            "Non-trainable params: 109,216\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YIBk0oTWkTAN",
        "outputId": "255f16c9-2c29-4b87-b4b8-ade7653ae1b1"
      },
      "source": [
        "print(len(model.layers))"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "408\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WViysE4Q2By5",
        "outputId": "87c4a959-8d09-486b-dc6d-1d3651296406"
      },
      "source": [
        "model.compile(optimizer = \"adam\" , \n",
        "               loss = 'sparse_categorical_crossentropy' , \n",
        "               metrics = ['accuracy'])\n",
        "# model.summary()\n",
        "\n",
        "callback1 = ModelCheckpoint(filepath='best_model.hdf5', save_best_only=True, save_weights_only=False)\n",
        "callback2 = EarlyStopping(monitor='val_loss', min_delta=0.1, patience=3, verbose = 1, mode='min', restore_best_weights = True)\n",
        "callback3 = ReduceLROnPlateau(\n",
        "    monitor = 'val_accuracy', \n",
        "    patience = 2, \n",
        "    verbose = 1, \n",
        "    factor = 0.3, \n",
        "    min_lr = 0.000001)\n",
        "\n",
        "history = model.fit(\n",
        "    train_images, \n",
        "    train_labels, \n",
        "    batch_size = 32, \n",
        "    epochs = EPOCHS, \n",
        "    validation_data=(val_images, val_labels), \n",
        "    callbacks=[callback1, callback3]\n",
        "    )"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "304/304 [==============================] - 58s 156ms/step - loss: 0.4512 - accuracy: 0.8369 - val_loss: 0.4669 - val_accuracy: 0.8230\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
            "  category=CustomMaskWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 2/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.2603 - accuracy: 0.9066 - val_loss: 0.2904 - val_accuracy: 0.9037\n",
            "Epoch 3/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.1761 - accuracy: 0.9338 - val_loss: 0.4957 - val_accuracy: 0.8469\n",
            "Epoch 4/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.1504 - accuracy: 0.9446 - val_loss: 0.2431 - val_accuracy: 0.9111\n",
            "Epoch 5/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.1375 - accuracy: 0.9505 - val_loss: 0.2305 - val_accuracy: 0.9078\n",
            "Epoch 6/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0872 - accuracy: 0.9696 - val_loss: 0.3221 - val_accuracy: 0.8905\n",
            "\n",
            "Epoch 00006: ReduceLROnPlateau reducing learning rate to 0.0003000000142492354.\n",
            "Epoch 7/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0369 - accuracy: 0.9865 - val_loss: 0.0749 - val_accuracy: 0.9745\n",
            "Epoch 8/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0274 - accuracy: 0.9901 - val_loss: 0.0716 - val_accuracy: 0.9737\n",
            "Epoch 9/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0255 - accuracy: 0.9905 - val_loss: 0.0737 - val_accuracy: 0.9695\n",
            "\n",
            "Epoch 00009: ReduceLROnPlateau reducing learning rate to 9.000000427477062e-05.\n",
            "Epoch 10/50\n",
            "304/304 [==============================] - 44s 143ms/step - loss: 0.0140 - accuracy: 0.9953 - val_loss: 0.0581 - val_accuracy: 0.9786\n",
            "Epoch 11/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0094 - accuracy: 0.9958 - val_loss: 0.0743 - val_accuracy: 0.9778\n",
            "Epoch 12/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0105 - accuracy: 0.9964 - val_loss: 0.0709 - val_accuracy: 0.9761\n",
            "\n",
            "Epoch 00012: ReduceLROnPlateau reducing learning rate to 2.700000040931627e-05.\n",
            "Epoch 13/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0077 - accuracy: 0.9965 - val_loss: 0.0682 - val_accuracy: 0.9786\n",
            "Epoch 14/50\n",
            "304/304 [==============================] - 44s 145ms/step - loss: 0.0072 - accuracy: 0.9971 - val_loss: 0.0673 - val_accuracy: 0.9786\n",
            "\n",
            "Epoch 00014: ReduceLROnPlateau reducing learning rate to 8.100000013655517e-06.\n",
            "Epoch 15/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0077 - accuracy: 0.9970 - val_loss: 0.0693 - val_accuracy: 0.9811\n",
            "Epoch 16/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0060 - accuracy: 0.9977 - val_loss: 0.0638 - val_accuracy: 0.9786\n",
            "Epoch 17/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0060 - accuracy: 0.9971 - val_loss: 0.0648 - val_accuracy: 0.9786\n",
            "\n",
            "Epoch 00017: ReduceLROnPlateau reducing learning rate to 2.429999949526973e-06.\n",
            "Epoch 18/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0056 - accuracy: 0.9974 - val_loss: 0.0669 - val_accuracy: 0.9786\n",
            "Epoch 19/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0069 - accuracy: 0.9973 - val_loss: 0.0667 - val_accuracy: 0.9753\n",
            "\n",
            "Epoch 00019: ReduceLROnPlateau reducing learning rate to 1e-06.\n",
            "Epoch 20/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0052 - accuracy: 0.9977 - val_loss: 0.0686 - val_accuracy: 0.9770\n",
            "Epoch 21/50\n",
            "304/304 [==============================] - 44s 145ms/step - loss: 0.0055 - accuracy: 0.9976 - val_loss: 0.0698 - val_accuracy: 0.9778\n",
            "Epoch 22/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0064 - accuracy: 0.9973 - val_loss: 0.0666 - val_accuracy: 0.9786\n",
            "Epoch 23/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0055 - accuracy: 0.9976 - val_loss: 0.0665 - val_accuracy: 0.9794\n",
            "Epoch 24/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0058 - accuracy: 0.9975 - val_loss: 0.0673 - val_accuracy: 0.9802\n",
            "Epoch 25/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0057 - accuracy: 0.9972 - val_loss: 0.0688 - val_accuracy: 0.9786\n",
            "Epoch 26/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0055 - accuracy: 0.9976 - val_loss: 0.0683 - val_accuracy: 0.9778\n",
            "Epoch 27/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0052 - accuracy: 0.9985 - val_loss: 0.0674 - val_accuracy: 0.9761\n",
            "Epoch 28/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0089 - accuracy: 0.9968 - val_loss: 0.0663 - val_accuracy: 0.9778\n",
            "Epoch 29/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0090 - accuracy: 0.9967 - val_loss: 0.0686 - val_accuracy: 0.9786\n",
            "Epoch 30/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0062 - accuracy: 0.9979 - val_loss: 0.0627 - val_accuracy: 0.9786\n",
            "Epoch 31/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0056 - accuracy: 0.9975 - val_loss: 0.0672 - val_accuracy: 0.9770\n",
            "Epoch 32/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0073 - accuracy: 0.9970 - val_loss: 0.0709 - val_accuracy: 0.9770\n",
            "Epoch 33/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0050 - accuracy: 0.9980 - val_loss: 0.0672 - val_accuracy: 0.9778\n",
            "Epoch 34/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0058 - accuracy: 0.9976 - val_loss: 0.0658 - val_accuracy: 0.9778\n",
            "Epoch 35/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0058 - accuracy: 0.9976 - val_loss: 0.0663 - val_accuracy: 0.9786\n",
            "Epoch 36/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0055 - accuracy: 0.9974 - val_loss: 0.0666 - val_accuracy: 0.9761\n",
            "Epoch 37/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0054 - accuracy: 0.9974 - val_loss: 0.0691 - val_accuracy: 0.9786\n",
            "Epoch 38/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0061 - accuracy: 0.9975 - val_loss: 0.0666 - val_accuracy: 0.9761\n",
            "Epoch 39/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0081 - accuracy: 0.9974 - val_loss: 0.0644 - val_accuracy: 0.9778\n",
            "Epoch 40/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0054 - accuracy: 0.9975 - val_loss: 0.0704 - val_accuracy: 0.9770\n",
            "Epoch 41/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0059 - accuracy: 0.9975 - val_loss: 0.0635 - val_accuracy: 0.9802\n",
            "Epoch 42/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0057 - accuracy: 0.9973 - val_loss: 0.0664 - val_accuracy: 0.9770\n",
            "Epoch 43/50\n",
            "304/304 [==============================] - 45s 149ms/step - loss: 0.0058 - accuracy: 0.9976 - val_loss: 0.0644 - val_accuracy: 0.9811\n",
            "Epoch 44/50\n",
            "304/304 [==============================] - 45s 149ms/step - loss: 0.0051 - accuracy: 0.9976 - val_loss: 0.0685 - val_accuracy: 0.9778\n",
            "Epoch 45/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0055 - accuracy: 0.9983 - val_loss: 0.0651 - val_accuracy: 0.9786\n",
            "Epoch 46/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0061 - accuracy: 0.9968 - val_loss: 0.0721 - val_accuracy: 0.9770\n",
            "Epoch 47/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0051 - accuracy: 0.9977 - val_loss: 0.0675 - val_accuracy: 0.9770\n",
            "Epoch 48/50\n",
            "304/304 [==============================] - 44s 144ms/step - loss: 0.0057 - accuracy: 0.9975 - val_loss: 0.0672 - val_accuracy: 0.9778\n",
            "Epoch 49/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0052 - accuracy: 0.9972 - val_loss: 0.0703 - val_accuracy: 0.9770\n",
            "Epoch 50/50\n",
            "304/304 [==============================] - 45s 148ms/step - loss: 0.0056 - accuracy: 0.9973 - val_loss: 0.0675 - val_accuracy: 0.9778\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BawWF4c5XvbA"
      },
      "source": [
        "def plot_accuracy_loss_chart(history):\n",
        "    epochs = [i for i in range(EPOCHS)]\n",
        "    fig , ax = plt.subplots(1,2)\n",
        "    train_acc = history.history['accuracy']\n",
        "    train_loss = history.history['loss']\n",
        "    val_acc = history.history['val_accuracy']\n",
        "    val_loss = history.history['val_loss']\n",
        "    fig.set_size_inches(20,10)\n",
        "    ax[0].plot(epochs , train_acc , 'go-' , label = 'Training Accuracy')\n",
        "    ax[0].plot(epochs , val_acc , 'ro-' , label = 'Validation Accuracy')\n",
        "    ax[0].set_title('Training & Validation Accuracy')\n",
        "    ax[0].legend()\n",
        "    ax[0].set_xlabel(\"Epochs\")\n",
        "    ax[0].set_ylabel(\"Accuracy\")\n",
        "\n",
        "    ax[1].plot(epochs , train_loss , 'g-o' , label = 'Training Loss')\n",
        "    ax[1].plot(epochs , val_loss , 'r-o' , label = 'Validation Loss')\n",
        "    ax[1].set_title('Training & Validation Loss')\n",
        "    ax[1].legend()\n",
        "    ax[1].set_xlabel(\"Epochs\")\n",
        "    ax[1].set_ylabel(\"Training & Validation Loss\")\n",
        "    plt.show()\n"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 604
        },
        "id": "4narSj_0XzZb",
        "outputId": "cd54474e-e4f5-4f1f-fe3c-4085b3caed0b"
      },
      "source": [
        "plot_accuracy_loss_chart(history)"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FE26pWCoX4Po",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "697f9b7e-7428-4cd7-d057-b8a5cb601cb2"
      },
      "source": [
        "results = model.evaluate(test_images, test_labels)\n",
        "\n",
        "print(\"Loss of test model is  \", results[0])\n",
        "print(\"Accuracy of test model is \", results[1]*100, \"%\")\n",
        "\n",
        "\n",
        "results = model.evaluate(val_images, val_labels)\n",
        "\n",
        "print(\"Loss of validation model is \", results[0])\n",
        "print(\"Accuracy of validation model is \", results[1]*100, \"%\")\n",
        "\n",
        "results = model.evaluate(train_images, train_labels)\n",
        "\n",
        "print(\"Loss of train model  is \", results[0])\n",
        "print(\"Accuracy of train model is \", results[1]*100, \"%\")"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "38/38 [==============================] - 2s 41ms/step - loss: 0.0327 - accuracy: 0.9885\n",
            "Loss of test model is   0.03268973529338837\n",
            "Accuracy of test model is  98.8477349281311 %\n",
            "38/38 [==============================] - 1s 38ms/step - loss: 0.0675 - accuracy: 0.9778\n",
            "Loss of validation model is  0.06754099577665329\n",
            "Accuracy of validation model is  97.77777791023254 %\n",
            "304/304 [==============================] - 12s 38ms/step - loss: 0.0043 - accuracy: 0.9983\n",
            "Loss of train model  is  0.004282092675566673\n",
            "Accuracy of train model is  99.82503056526184 %\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L0pT58OzX-sL",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ef4eaa59-28d5-4ccc-8a3e-d7884b1f8898"
      },
      "source": [
        "from sklearn.metrics import classification_report\n",
        "\n",
        "predictions = model.predict(test_images)\n",
        "predictions = np.argmax(predictions,axis=1)\n",
        "predictions[:15]"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0, 3, 1, 1, 0, 2, 2, 3, 1, 2, 1, 2, 1, 3, 1])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VcIppIdzYAz8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0d80e3d7-f8ba-413e-dcf2-56bbb1f911c0"
      },
      "source": [
        "print(classification_report(\n",
        "    test_labels, \n",
        "    predictions, \n",
        "    target_names = ['EOSINOPHIL (Nhom 0)', 'LYMPHOCYTE (Nhom 1)', 'MONOCYTE (Nhom 2)', 'NEUTROPHIL (Nhom 3)']))"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "                     precision    recall  f1-score   support\n",
            "\n",
            "EOSINOPHIL (Nhom 0)       0.98      0.98      0.98       291\n",
            "LYMPHOCYTE (Nhom 1)       1.00      1.00      1.00       301\n",
            "  MONOCYTE (Nhom 2)       0.99      0.99      0.99       308\n",
            "NEUTROPHIL (Nhom 3)       0.99      0.99      0.99       315\n",
            "\n",
            "           accuracy                           0.99      1215\n",
            "          macro avg       0.99      0.99      0.99      1215\n",
            "       weighted avg       0.99      0.99      0.99      1215\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cdJWGyhJYEoG",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 166
        },
        "outputId": "62247938-f5ff-4fef-b793-f4be946f5d5b"
      },
      "source": [
        "cm = confusion_matrix(test_labels, predictions)\n",
        "cm = pd.DataFrame(cm, index = ['0', '1', '2', '3'], columns = ['0', '1', '2', '3'])\n",
        "cm"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>0</th>\n",
              "      <th>1</th>\n",
              "      <th>2</th>\n",
              "      <th>3</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>285</td>\n",
              "      <td>0</td>\n",
              "      <td>4</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0</td>\n",
              "      <td>301</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3</td>\n",
              "      <td>0</td>\n",
              "      <td>304</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>4</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>311</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "     0    1    2    3\n",
              "0  285    0    4    2\n",
              "1    0  301    0    0\n",
              "2    3    0  304    1\n",
              "3    4    0    0  311"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 38
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WovW0OfFHOV3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 592
        },
        "outputId": "1eaad673-fc88-4fb4-b9fc-933f70175271"
      },
      "source": [
        "def plot_confusion_matrix (cm):\n",
        "    plt.figure(figsize = (10,10))\n",
        "    sns.heatmap(\n",
        "        cm, \n",
        "        cmap = 'Blues', \n",
        "        linecolor = 'black', \n",
        "        linewidth = 1, \n",
        "        annot = True, \n",
        "        fmt = '', \n",
        "        xticklabels = class_names, \n",
        "        yticklabels = class_names)\n",
        "    \n",
        "plot_confusion_matrix(cm)"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}