README.md

Deep-Learning-Projects

This repo contains projects related to Vision, NLP and Reinforcement Learning

Building from Scratch ❄️

1. Sentiment analysis in numpy 😊 😕

Building a neural network from scratch. We understand the effect of importance of learning rate and input size on training speed. Also we check out the learnt embeddings of similar words

2. Effect of different weight initializations 🔮

Here we discuss different methods of weight initializations and also dicuss why things work the way they do.

Natural language Processing 🌟

1. Loading a custom Text dataset 🎶

Loading a custom dataset in pytorch for machine translation. Can be used for both text and images.

2. Sequence to Sequence model for Machine Translation 🎏

An Encoder Decoder based model to translate from English to Hindi. The code is customizable to be used for other translation tasks.

3. Quadratic Solver Transformer 🔥

And end-to-end Encoder Decoder transformer project to solve Quadratic Equations. Also has code for attention maps

4. Multi-Task-Learning 🎯

Learning NER tags and Intent joinly with the same model

Computer Vision 👀

1. Landmark Classification 🗼

Classify images into 50 Landmarks. Albumentations for augmentation. And a pretrained Resnet-34 for transfer learning

2. Loading a custom image dataset 🧁

Loading a custom dataset in pytorch for image classification

Reusable Codes 🌀

1. Hooks 🎯

1.1 Forward Hooks Learn about hooks in pytorch (Youtube)

Forward hooks can be used for model debugging and create activation maps

2. Resume Training from Checkpoint ▶️

This code saves training information into checkpoints and use it to resume training