Our paper “Depth-Assisted Sparse Visual Odometry for UAV-Relevant Synthetic RGB-D Evaluation: A Controlled Geometric-Backend Ablation” is now published in Robotics.

The work studies a controlled sparse visual odometry backend for robotics-oriented synthetic RGB-D evaluation. The main question is practical: when depth is available, how much does it help a geometric visual odometry pipeline, and where does the improvement come from?

We use synthetic RGB-D data to compare backend variants under controlled conditions. This makes it easier to separate the effect of depth-assisted geometry from unrelated dataset or implementation noise. The result is a focused ablation of sparse VO design choices for robotic perception research, with UAV navigation as one important application context.

Links: