A Fully Automated open-source AI-powered agentic framework which builds your startup with AI Employees like a real company.

repo to house models to perform real estate image manipulaton

microsoft ai hackathon submission for cloud coverage segmentation

Machine Learning model to segment floodwater regions from satellite (Sentinel-1) radar imagery

3D Point Cloud Classification

A lightweight ConvNet (~700 KB) to classify pictures of different rooms of a house/apartment with 88.9 % accuracy

Maze Solver from scratch (under 260 lines) using Naive Reinforcement Learning with Q-Table construction

repository containing rough code for deep reinforced control of autonomous vehicles

pythonic codebase to do 3D reconstruction stuff

A fast, purely Numpy based library implementing Fully Connected and Convolutional Neural Nets

Codebase to perform real-time detection+tracking of cars across different camera sources

Pruned Lightweight LeNet using 4% weights to acheive 90% top-1 accuracy on MNIST

A C++ map matching application which matches probe GPS data with road link data. This is used to align noisy GPS data with actual road networks.

Exploratory attempt to implement Keypoint based 3D reconstruction from scratch in C++/CUDA

Parallel GPU Implementation of Connected Component Labelling (CCL). Connected-component labeling is used in computer vision to detect connected regions in binary digital images

Windows Presentation Foundation (WPF) based application developed for the career services cell of JK Institute of Applied Physics, University of Allahabad

Deep Learning Powered Arrhythmia (Heart Disease) Detector

A CUDA & cuDNN powered Deep Learning C++ Library with clean interface and faster training speed (6x more over Caffe)

A neat C++ custom Matrix class to perform super-fast GPU (or CPU) powered Matrix/Vector computations with minimal code, leveraging the power of cuBLAS where applicable.

Application to generate projections of 3D point-cloud images using CUDA and computation of subsequent misalignment with its corresponding 2D images

This is a fast custom algorithm having O(n) linear time & O(n) memory complexity implemented on the CPU for solving the famous connected component labelling problem. The algorithm implemented here,…

A C++ simulator to simulate Link State Routing protocols used in Computer Networks

Python implementation of a Computer Vision algorithm to automatically extract lens smear regions from a batch of input images.