3DObjectTracking

Why we included this project

Researchers working on robot manipulation or augmented reality often need to track objects in 3D space, and this repository from the DLR robotics group collects several published 6DoF tracking algorithms, each in its own folder with the code needed to reproduce the paper's results. The methods are built for the messy conditions of real tracking: partial occlusion, motion blur, background clutter, and textureless objects, combining region information with depth and appearance cues. The latest version, M3T, goes further by handling multi-body kinematic structures, which is useful when a robot has to manipulate articulated things like grippers or tools. Since each release updates earlier work, you can jump straight to the newest folder instead of sorting through older implementations. It's a research codebase rather than a ready-made library, so expect to adapt C++ code to your own pipeline, but the reference implementations and their papers give you a solid foundation for building RGB-D trackers.