PROJECTS

A program to detect people who are not adhering to a the social distancing norm.

I built a custom GAN which can generate images of clothes. The GAN was trained on 937 images from the MNIST fashion dataset.

A Fully Convolutional Network that can segment roads in satellite Images.

A program that uses CNN to classify facial emotions into one of the five basic emotions - Fear, Anger, Disgust, Sadness, and Happiness.

Using sentiment analysis to estimate the happiness of cities by analyzing the location-enabled tweets.

This project involves semantic segmentation of Indian rooftops from low-resolution satellite images.

Forecasting air pollution levels using a simple RNN.

Used SOTA Face Recognition Models to create an identity recognition and tracking model to mark attendance in a medium-sized organization.

Counting People in Dense crowds is a laborious task, but this LSC-CNN implementation is the perfect solution which gives an accurate headcount.

A face recognition model that can be used to recognize custom faces. I trained it to recognize faces from the famous sitcom 'Friends'.

A CNN which is capable of recognizing characters from the English lexicon, in images.

A CNN that can recognize fruits in an image with a decent accuracy.

Using Sentiment analysis to analyze feedback from students. This can be used to gather and analyze large volumes of feedback.

A project that explores diverse sentiment analysis models on live tweets.

Detecting and reporting incidents using advanced objects detection and tracking to observe vehicles on Highways. Currently deployed at multiple toll stations in Northern India.

A Unet to segment nuclear region in cellular images.

This project employees computer vision and Reinforcement learning to simulate eco-friendly and sustainable city growth. This project is a work in progress.

A CNN to recognize gestures in real-time. This project employees basic computer vision techniques to isolate the hand and then use a neural network to predict the gesture.

A program that can count the number of squats performed by a human subject through a webcam feed. This program uses Pose-net to accurately count the number of squats.

A CNN that can predict whether an image contains a cat or a dog. This was one of my first deep learning projects.