Reproduce “FreeOcc: Decoupling Ego-Motion for Efficient 4D Occupancy Forecasting via Continuous Flow Matching” 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
Zeping Zhang, Zhuoya Zhao, Samy Metari, Robert Laganière
FreeOcc forecasts future three-dimensional road occupancy by geometrically warping bird’s-eye-view features with the vehicle’s future pose, then predicting residual scene changes through decoder and continuous-time flow-matching paths blended by uncertainty. On nuScenes, its ensemble reaches 32.1% average class-agnostic voxel mIoU across six horizons, versus 28.9% for OccWorld; planar warping assumes roughly flat roads.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.