ocr_hog+lreg.py

# -*- coding: utf-8 -*-
"""OCR HOG+LREG.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1YVf4pmUbLUT4kOyII2cjc0L6GHIWR4Jx
"""

import numpy as np
import os
from skimage.feature import hog
import joblib
import cv2
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
import pandas as pd

# train_dir = r'/content/drive/MyDrive/YourFolderName/'
train_dir = r'ocr_data/train/'
labels_list = [i for i in os.listdir(train_dir)]
pathimg = [os.listdir(train_dir + i) for i in labels_list]

# Visualize HOG for letter A
im_test =  cv2.imread('ocr_data/train/A/roi107644.jpg',0)
_,hog_img= hog(im_test,orientations=9,pixels_per_cell=(8,8), cells_per_block=(1, 1),visualize=True)
plt.imshow(hog_img,cmap='gray')
# cv2.imwrite('AHog.jpg',hog_img)

# extract the hog for each image and store it in a list with its 
# corresponding label
features = []
labels = []
for i,j in enumerate(zip(pathimg,labels_list)):
  imgs,label = j
  for img in imgs:
      img = cv2.imread(train_dir+label+'/'+img)
      img_res=cv2.resize(img,(64,128),interpolation=cv2.INTER_AREA)
      img_gray= cv2.cvtColor(img_res,cv2.COLOR_BGR2GRAY)
      hog_img= hog(img_gray,orientations=9,pixels_per_cell=(8,8), cells_per_block=(1, 1))
      features.append(hog_img)
      labels.append(label)

print(len(pd.DataFrame(np.array(features))))
print(len(pd.DataFrame(np.array(labels))))

df = pd.DataFrame(np.array(features))
df['target'] = labels
df

df['target'].unique()

# df.target.value_counts()
sns.countplot(x='target', data=df)

"""#Training"""

x = np.array(df.iloc[:,:-1])
y = np.array(df['target'])

from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
x_train, x_test, y_train, y_test = train_test_split(x, y,
                                                    test_size=0.20,
                                                    random_state=42)

from imblearn.over_sampling import SMOTE
sm = SMOTE(random_state=0)
sm_x,sm_y=sm.fit_sample(x_train, y_train)

bal_df = pd.DataFrame(sm_x)
bal_df['target']=pd.DataFrame(sm_y)
sns.countplot(x='target', data=bal_df)

bal_df['target'].value_counts()

lreg = LogisticRegression()
clf=lreg.fit(sm_x, sm_y)
y_pred = clf.predict(x_test)
print('Accuracy {:.2f}'.format(clf.score(x_test, y_test)))

from sklearn.metrics import classification_report
print(classification_report(y_test, y_pred))

# save trained model
joblib.dump(clf, r'ocr_data/models/hog_lreg_model_3.pkl')

from google.colab.patches import cv2_imshow

"""#Testing"""

# Load the classifier
clf = joblib.load("ocr_data/models/hog_lreg_model_3.pkl")

# Read the input image
im = cv2.imread("ocr_data/licenseplates/licplate4.jpg")

# Convert to grayscale
im_gray = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)

# Threshold the image in order to find contours
ret, im_th = cv2.threshold(im_gray, 120, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)

# Find contours in the image
ctrs,hier = cv2.findContours(im_th, cv2.RETR_TREE,
                                 cv2.CHAIN_APPROX_SIMPLE)

bboxes=[cv2.boundingRect(c) for c in ctrs]
sorted_bboxes = sorted(bboxes,key=lambda b:b[0])

# For each rectangular region, calculate HOG features and predict
# the digit using our logistic regression model.
plate_char=[]
for num,i_bboxes in enumerate(sorted_bboxes):
    [x,y,w,h]=i_bboxes
    if h>100 and w < 100:
    # Make the rectangular region around the digit
        cv2.rectangle(im,(x,y),(x+w,y+h),(0,255,0),1)
        roi=im_gray[y:y+h,x:x+w]
        # Resize the image
        roi = cv2.resize(roi, (64, 128), interpolation=cv2.INTER_AREA)
        # Calculate the HOG features
        # use the same parameters used for training
        roi_hog_fd = hog(roi, orientations=9, pixels_per_cell=(8, 8),
                         cells_per_block=(1, 1))
        nbr = clf.predict(np.array([roi_hog_fd]))
        
        cv2.putText(im, str((nbr[0])), (x,y+h),cv2.FONT_HERSHEY_SIMPLEX,
                    2, (0, 200, 250), 3)
        plate_char.append(str(nbr[0]))

print(''.join(plate_char))
                     
cv2.imshow('result',im)