DytanVO: Joint Refinement of Visual Odometry and Motion Segmentation in Dynamic Environments
This repository will be used to release our code for the most recent ICRA 2023 submission, DytanVO: Joint Refinement of Visual Odometry and Motion Segmentation in Dynamic Environments. DytanVO is the first supervised learning-based VO method that handles dynamic environments.
(For the archived codebase Dynamic Dense RGB-D SLAM with Learning-Based Visual Odometry, please check out the legacy branch)