Implementation of the paper KinectFusion by Microsoft using C++ and CUDA. The goal of the project is to develop a method for accurate real-time mapping of indoor scenes, using only a low-cost depth camera and graphics hardware. Presentation video can be seen below.

Implemented vanilla Tracktor with Faster-RCNN using PyTorch. Added a basic motion model and also a reidentification network as the paper suggested for Tracktor++.

We extended the paper “Divergence-Free Shape Correspondence by Deformation” and representing a motion sequence as a time dependent vector field. We solve correspondence and matching problems for the whole sequence during the optimization. Some of the current results can be seen below. …