Abstract: Recently, video recognition is emerging with the help of multi-modal learning, which focuses on integrating distinct modalities to improve the performance or robustness of the model.

Abstract: Visual odometry (VO) is a key part of autonomous navigation systems, particularly for robots and autonomous vehicles. Conventional feature-based or direct approaches for VO are powerful but ...