Reproduce “RCBEVDet++: Toward High-accuracy Radar-Camera Fusion 3D Perception Network” 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

Zhiwei Lin, Zhe Liu, Yongtao Wang, Le Zhang, Ce Zhu

RCBEVDet++ improves camera–radar 3D detection by encoding sparse radar returns into denser bird’s-eye-view features and aligning radar and camera features with cross-attention before fusion. It evaluates detection, segmentation, and tracking on nuScenes. On nuScenes’ test set, its V2-99 camera-plus-radar model reports 68.7 NDS and 62.6 mAP, versus 63.6 NDS and 55.6 mAP for same-backbone, camera-only SparseBEV; neither uses future frames.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.