Reproduce “MapLightning: Online Vectorized HD Map Construction with 1D Map Tokens” 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

Shen Zheng, Anurag Ghosh, Mani Ramanagopal, Srinivasa Narasimhan

MapLightning replaces dense bird’s-eye-view grids for online map construction with learned map tokens that gather information from multi-camera image features through self-attention before a decoder produces vectorized road elements. In a single-frame ResNet-50 nuScenes comparison, it achieves 30.1 mAP versus 25.5 for cross-attention using the same 0.3k tokens. These results concern map construction on nuScenes and Argoverse 2.

Summary by alphaXiv.

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