Reproduce “HorizonStream: Long-Horizon Attention for Streaming 3D Reconstruction” 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
Chong Cheng, Peilin Tao, Nanjie Yao, Guanzhi Ding, Xianda Chen, Yuansen Du, Xiaoyang Guo, Wei Yin and 4 more
HorizonStream performs streaming 3D reconstruction by matching nearby frames and carrying a fixed-size geometric state whose channels retain evidence at different learned rates. Metric readout tokens use this state to help recover scale. Across seven full-length VBR sequences, average ATE is 25.30 m versus 27.53 m for Lingbot-map without loop closure; Lingbot-map has lower ATE on two sequences.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.