NeRF: Neural Radiance Fields

Object Detection and Instance segmentation approaches

Implementation of Various Object Detection and Instance Segmentation Algorithms such as YOLO and SOLO

Autonomous Pick and Place using
7 DOF Robot

Autonomous control of 7-DOF Franka Emika Panda Robot Arm for picking and stacking stationary and moving blocks

Motion Planning and Trajectory Generation for Quadrotors

Implementated non-linear geometric controller, Error-State Kalman filter for estimating the pose and orientation of the quadcopter based on inputs from stereo cameras and the IMU. Implemented a Minimum Snap Trajectory Generation Algorithm using sparse points from the path using A* Algorithm and Ramer Douglas Peucker Algorithm.

Visual Odometry using
RGB-D Images

Estimated the 6-DoF trajectory of a moving camera by aligning sparse features obtained from RGB-D images against a persistent model set consisting previous features. A Kalman Filter is used to dynamically update the model set by incorporating new features obtained from each RGB-D frame. A Gaussian mixture model is implemented to handle the uncertainty in the sparse RGB-D image features.

3d Reconstruction from Images

3D Point Cloud reconstruction using Two-View and Multi-View Stereo. Generated the disparity map using distance metric kernels such as ssd, sad and zncc which gave the depth map of the scene in Two-View and constructed a cost volume by stacking depth cost maps obtained by projecting reference view and warped neighboring view using homography in the case of multi view.