Deep Learning

This repository is an ensemble of different kinds of applications implemented using Tensorflow and Keras. A very good guide for beginners. The following table gives a overview(besides the name) of what different folder contains.

Table of contents:

Folders	Descriptions
DeepNetBasic	Low level implementation of shallow/basic feed forward network to perdom classification task on MNIST dataset using custom training loop.
DenseNet	Implementation of DenseNET network to perform classification task on CIFAR10 dataset, and experiments with different hyper-parameters.
ExperimentsWithOptimizers	Experiments with different regularizer choices(and different parameters) and their impact on model training.
ExperimentsWithRegularizers	Experiments with different optimizer choices(and different parameters) and their impact on model training.
Autoencoder_basic	Basic Autoencoder implementation to reconstruct the MNIST dataset digits. The notebook also contains an small experiment of walking the code(latent) space.
CatsVsDogs	Binary classification between cats and dogs images from Kaggle dogs-vs-cats dataset, using RESNET architecture. The model is inspired from Coursera course.
HandGestureDigitRecog	Classification of digits from images, where digits are shown by hand gesture, using LENET model.
Titanic - Kaggle	Binary classification task performed on Titanic dataset from Kaggle using (not so deep) Dense model using Keras. Good example for ultimate starters.
RNN_LowLevel	Low level implementation of RNN basic concepts for character level data generation.
LSTM	Implementation of RNN using LSTM(Tensorflwo) for character level sequence generation. The model is capable of taking variable length input sequence.
NMTWithAttention	Attention based Neural Machine Translation(NMT) implemented to translate from bengali to english. This notebook implements and compare three different attention mechanisms- Bahdanau, Dot-product & Luong product.
ObjectDetectionInImage_YOLOv3	Object detection task perform using YOLOv3 model on images. Code and model are inspired by Coursera course.
ObjectDetectionInVideo_YOLOv3	Object detection task perform using YOLOv3 model on video frames. Code and model are inspired by Coursera course.
TransferLearning	This notebook shows implementation of different transfer learning strategies and their comparison.
BERTFineTuning	A BERT model is fine-tuned with a binary classification task of finding grammatical correctness of a sentence. This implementation uses 'glue-CoLA' dataset.
VAE	This notebook contains implementation of a basic Variational Autoencoder(VAE) on MNIST dataset.
ImprovedGAN_WGAN	Implementation of Wasserstein GAN(WGAN) and WGAN with Gradient Penalty(WGAN-GP) to generate new images for FMNIST dataset.

Note:

Some of the implementations/experminets are inspired by some blogs. You can find relevant links in the notebooks itself.

---

Citation:

If this repository helps you or you use some of its code, please refer the following archive as citation in your codes/publications.

Contact:

For any issues, clarification, bug or improvement suggestion, please send an email to arnabdas8901@gmail.com. I will reach back as soon as possible.