Monocular Visual Odometry

RBCCPS, IISc

• Research project to explore the challenges and applications of Visual Odometry.
• Compared deep-learning-based feature extractors and traditional features for indoor and outdoor navigation tasks.
• Identified bottlenecks in the VO pipeline and explored learned feature descriptors for enhanced performance.
• Published a comprehensive survey on advancements in monocular visual odometry, and used the findings to design an emergency autonomous parking application for Advanced Driver Assistance Systems (ADAS). See publications for the paper.