Reproduce “MotoVerse: A Benchmark for Predictive Perception of Motorized Two-Wheelers in Dense Unstructured Traffic” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Varun A. Paturkar, Shankar Gangisetty, C.V. Jawahar
MotoVerse benchmarks single-camera prediction of motorized two-wheelers’ maneuvers and nearby agents’ motion. It includes 949 maneuver clips and trajectories for over 31,000 agents. Video models retain 84–93% of full-clip accuracy from the first fifth, though early accuracy remains near 50%. For trajectory prediction, SGNet achieves 51.8-pixel ADE versus 94.3-pixel for a static baseline in image-plane error, not safety outcomes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.