DL_Training

Task list

1. Linear Regression
   1. Backpropagation
   2. Feature scaling
   3. Gradient descent
   4. Quadratic Regression (Optional)
2. Neural Networks (ANN)
   1. Activation Functions
   2. Types of activation functions
   3. Different types of gradient descent
   4. Dropout layers
   5. Optimizers and its types
3. CNN(Convolution neural networks) –
   1. Convolution
   2. padding
   3. Max pooling
   4. Skip connections
   5. VGG fine tuning
4. FCN (Fully convoluted Neural Networks)
   1. Basic implementation
   2. FCN-DenseNet implementation
5. Encoder-Decoder
   1. Basic implementation
   2. UNet implementation
6. Transformer(optional)
7. GANs (optional)

Resources

Theory

1. 3Blue1Brown - Introduction to Neural Networks
2. Deep learning with Pytorch - Book
3. KrishNaik - Deep Learning Playlist
4. Andrew NG - Deeplearning.ai 5 Playlists of specialisation
5. MIT 6.S191 - Intro to deep learning
6. Pytorch Tutorials
7. For basic intuition (You can find torrent) - Deep Learning A-Z™: Hands-On Artificial Neural Networks
8. Deep Learning book by Ian Goodfellow
9. DS-GA 1008 · SPRING 2020 by Yann LeCun

Practise (Blogs and Notebooks) with Video explanation in some

1. Pytorch Fundamentals by Microsoft learn
2. Going from Tensorflow to Pytorch
3. Notebooks with video explanation - PyTorch for Deep Learning by Jovian.ai
4. Focus on Pytorch Projects only - Mithi DL video and projects

Problem Statement

1. Week 1 - you have to use a DenseNet for Semantic Segmentation on COCO dataset. Structure of DenseNet which has to be used is given in this link. You can find the implementation of this paper here. Coco dataset is a very famous dataset used for segmentation purposes. it can be found in kaggle and it is also part of torchvision library. If you are having any storage issues with COCO dataset, you can also use MiniCOCO dataset. Solution