Reproduce “FastBEVFusion: Simple Multi-Modal Fusion for 3D Object Detection at 50 FPS” 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
Tom Sanitz, Emre Gülsoylu, Simone Frintrop, Christian Wilms
FastBEVFusion combines camera and LiDAR features for nuScenes 3D object detection without estimating exact image-feature depth. It spreads image features along viewing rays and uses a single cross-attention block to match them with LiDAR features. On nuScenes validation, its ResNet-18 version reports 62.1 mAP and 65.9 NDS at 17 ms per frame on an RTX 4090.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.