Reproduce “USI-Brain: Enhancing Underwater Spatial Intelligence via Spatial Data-Driven Vision-Language Modeling” 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

Zhiliang Zhu, Zhiqiang Zhao, Xie'an Jiang, Yingxin Lai, Kangye Zheng, Dawei Zhao

USI-Brain trains an underwater vision-language model to answer spatial questions by linking enhanced images with object-aligned depth and reconstruction cues. Its data engine uses these cues to generate a shared question-and-answer corpus, while the model receives raw images at evaluation. On a held-out benchmark, it improves the mean signed score difference by 5.76 points, or 7.63 with UPR, though some tasks remain below the backbone.

Summary by alphaXiv.

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