aju22

DocumentGPT

DocumentGPT is a web application that allows you to chat over your research document using OpenAI's chat API and perform semantic search using vector databases. This tool provides a seamless interface for interacting with your research document, exploring search results, and engaging in a conversation with an AI chatbot.

LLaMA2

This repository contains an implementation of the LLaMA 2 (Large Language Model Meta AI) model, a Generative Pretrained Transformer (GPT) variant. The implementation focuses on the model architecture and the inference process. The code is restructured and heavily commented to facilitate easy understanding of the key parts of the architecture.

DDPM

This is an easy to understand, simplified, broken-down implementation of Diffusion Models written in PyTorch. The architecture is borrowed from the paper "Denoising Diffusion Probabilistic Models"

RoPE-PyTorch

This repository contains an educational implementation of Rotary Positional Encodings (RoPE) in PyTorch. RoPE is a method introduced in the paper RoFormer: Enhanced Transformer with Rotary Position Embedding by Jianlin Su, Yu Lu, Shengfeng Pan, Ahmed Murtadha, Bo Wen, and Yunfeng Liu.

DQN-2048

Implementation of Deep Q-network to play game 2048 using Tensorflow2.0/Keras and PyGame

Real-Time-ISL-Translation

Uses deep learning to translate Indian Sign Language in real-time

Transformer-Distribution-Visualizer

A Research Toolkit that allows users to select a transformer model, specify a decoding strategy to generate text, which are then displayed in an interactive fashion to understand model probabilities.

Sudoku-Bot

An interactive python program to solve sudoku.

VQ-GANs

This is a simplified implementation of VQ-GANs written in PyTorch. The architecture is borrowed from the paper "Taming Transformers for High-Resolution Image Synthesis".

Enhanced-SRGAN

This is an implementation of ESRGAN's in Tensorflow2.0/Keras.

GoogleNet

This is an implementation of GoogleNet/InceptionNet from scratch, in Tensorflow-Keras

GPT2-From-Scratch

A ground-up implementation of the GPT2 model

Graph-Convolutional-Network-for-ESOL

This is an implementation and explanation of Graph Neural Networks for predicting water solubility of small organic molecules based on their chemical composition and structural information in PyTorch Geometric

ML-From-Scratch

This repository contains Python implementations of various machine learning algorithms from scratch using only native Python and NumPy.

Object-Tracking-OpenCV_Arduino

This project operates a pan-tilt-servo camera mount by an Arduino UNO to track any object.

YOLO-v1

This is an implementation of Yolo v1 paper built from scratch using Tensorflow-Keras with other utilities for working with the model and generating custom data.

Transformers

Implementation of Transformers from scratch given by the paper "Attention is all you Need" in Tensorflow/Keras.

aju22

AugViz

AugViz provides an easy-to-use graphical interface to visualize various augmentations on images in real-time.

DeepFROST-Remembering-Robert-Frost

This project uses LSTMs and Pre-Trained WordVecs in Tensorflow-Keras, to generate Frost-like poetry.

EfficientNet

This is a model implementation of EfficientNet(including all the versions from b0-b7), using Tensorflow-Keras.

Genetic_Algorithm

Visualising Genetic Algorithm, with a bunch of rockets.

Pi_Collision

Estimating Digits of Pi using Physics Collisions.

pix2pix

This is a model architecture implementation of Image-Image Translation pix2pix model, from scratch using Tensorflow-Keras.

Potrait-Segmentation-BodyPix

This project is a semantic background-foreground segmentation for human bodies using BodyPix to convert your photos to have Bokeh effect or to insert yourself in a virtual background.

pytorch-tutorial

PyTorch Tutorial for Deep Learning Researchers

ROS-Navigation-Stack-and-SLAM-for-Autonomous-Custom-Robot

This is workspace used in the the course "Building a bot in from scratch in ROS "

Seq2Seq-Math-Addition

This is a Seq2Seq model for predicting addition of two numbers as a sequence learning problem in Tensorflow/Keras.

UNet

This is an implementation of UNet Architecture for Semantic Segmentation from scratch, using Tensorflow-Keras

Variational-Auto-Encoders

This is a broken-down simplified implementation of Variational Auto Encoders in Tensorlflow2.0/Keras. Also includes a toy training example and image synthesis from an anime face dataset.