Projects

Multi-Mesher

A Diffusion Driven single 2D image to 3D Mesh Reconstruction using Zero123 and PointNet architecture

Masked Autoencoder Inpainting

A Transformer-based Autoencoder for Collaborative Perception by Image Inpainting

American Sign Language Detection

A CNN and LSTM based American Sign Language (ASL) Detector for Letters from video feeds

Volume Rendering and NeRFs

Implementation of Neural Radiance Fields (NeRF) for 3D Reconstruction and Novel View Synthesis

Classifying & Segmenting Point Clouds

A PointNet based architecture for classification and segmentation in pointclouds

Pedestrians Future Path Prediction

A CNN and bi-directional LSTM based pedestrian detection, tracking and future path prediction pipeline

Panoramic Image Stitching

Stitching multiple images together to create a panoramic view using homography techniques

Leaf Detection

A YOLO based Leaf Detection pipeline to aid plant health monitoring for farm robots

Stereo Vision System

Implementation of a stereo vision system using two calibrated cameras to predict the depth of objects in different scenes

Camera Calibration

Implementation of Camera calibration, using Direct Linear Transform and by using out-of-the-box OpenCV functions

Camera Pose Estimation

Implementation of pose estimation of a camera observing a planar object using homography techniques

RaspberryPi based mobile robot - TerpBot

A RaspberryPi based Unicycle Mobile Robot using wheel odometry and monocular perception, built from scratch

Planning with RRT* Smart & Potential Fields for TerpBot*

Global path planning using A* and local dynamic path planning using potential fields algorithm for TerpBot

Leader-Follower using TerpBot*

TerpBot following a human leader along with dynamic obstacle avoidance using a synthetic 2D point cloud generated with MiDaS Monocular Depth Estimation

Transformer Robot for Picking & Placing

End-to-end modelling of a mobile manipulator integrated with ROS package for joint control to simulate an autonomous pick & place task in Gazebo

Turtlebot Aruco based Maze Completion

C++ based ROS package which makes a TurtleBot3 complete a maze by detecting and using aruco markers' information in Gazebo

Turtlebot Map Navigation and Parts Search

C++ based ROS project which makes a TurtleBot3 navigate a map in Gazebo using logical cameras and search for parts spread out in the map

A* Path Planning

A* Path Planning algorithm implementation for a Mobile Point Robot and a TurtleBot3

Toycar Modeling

End-to-end modelling of a toycar and teleoperational control of the same

Dijkstra Path Planning

Dijkstra Planning algorithm implementation for a Mobile Point Robot

Voice Anti-Spoofing

A lightweight Support Vector Machine (SVM) based anti-spoofing model which can distinguish between audio from a real human and a playback device

PPO, DQN and Dueling DQN for Kuka Pick & Place

Implementation of PPO, DQN and Dueling DQN RL algorithms for Kuka pick and place task in PyBullet using OpenAI gym

Playing Atari with Deep Reinforcement Learning

Re-implementation of the famous Deep Reinforcement Learning paper from scratch for playing multiple Atari games

Hostage Rescue Game

A Virtual Reality game based on unity engine which simulates a hostage rescue mission

Exploring Immersive Locomotion Interfaces in Virtual Reality

Implementation of different common and uncommon VR locomotion interfaces

Creating Immersive VR Avatars with Animation Rigging

Implementing inverse kinematics in unity engine to create realistic VR avatars

Computer Vision