Reproduce “DepthEvidence: Unifying Metric Depth Prediction and Geometric Reasoning in Multimodal Language Models” 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

Jiangning Wei, Yuan Yao, Miaomiao Cui, Mingsheng Li, Humen Zhong, Shuai Bai, Zhibo Yang

DepthEvidence connects dense metric-depth prediction to language reasoning by encoding object-aligned geometry tokens tied to referenced object IDs. It pools predicted depths and decoder features for each object, with supervision to preserve depth information. On Depth-VQA, it raises Relative Overall Accuracy from 0.5838 to 0.6270 and Metric Overall Accuracy from 0.7692 to 0.8504; evaluation assumes camera intrinsics and grounded references.

Summary by alphaXiv.

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