Spot Spraying My project at DFKI with HYDAC. Built a robotic system for identifying & localizing weeds/crops for spot-spraying. More about the project is can be found on this Blog.

DeepLiDARFlow : Bachelor thesis; worked on scene flow estimation using monocular camera & sparse LiDAR, work from the thesis has been accepted at two conferences. Papers & the corresponding code can be found in the publications page and the link below. Work was completed while working at Augmented Vision, DFKI, Kaiserslautern, Germany.
[Code]

Unsupervised Cross Spectral Stereo Matching PyTorch implmentation of the paper [Unsupervised Cross-spectral Stereo Matching by Learning to Synthesize] The method uses CycleGAN to learn translation from one spectrum to the other and then uses it to train a stereo matching network based on unsupervised loses. Currently working on integrating contrastive losses for robust image translation.
[Code]

Generative Adversarial Networks for Visual Odometry: Unofficial pytorch implementation for GANVO - it uses visual odometry for estimating depth. This work remains one of the very few GAN based work on visual odometry.
[Code]

HSID-CNN: Worked on Deep Learning for cleaning and extraction of useful bands in hyperspectral images, Implemented HSID-CNN in tensorflow for denoising hyperspectral images (AVIRIS). Work was completed while working at Pixxel, Bengaluru.
[Code]

Compiler Construction: Constructed a compiler for a given language specification in C language, this included the development of lexer, parser, semantic-analyzer, code-generator. This project was a part of Compiler Construction course at BITS.
[Code]

Visual Commonsense Reasoning: Implemented VCR in tensorflow. Used a new attention mechanism based--the results improved marginally by about 5%. This project was a part of my last semester project. Pandemic has stalled the organization of this codebase :P.

Active Learning: Implemented active learning algorithms on MNIST, tested various techniques like Query by committee & uncertainty sampling, also tested cluster based testing technique where whole dataset was labelled on the basis of just 10% of points. This project was a part of Machine Learning course at BITS.
[Code]

Structural Health Monitoring:Implemented auto-regressive model for feature extraction and several unsupervised algorithms like one-class SVM for final classification. Work was completed while working at CEERI, Pilani.
[Code]