Reproduce “IntriAdapt: Generalizable Intrinsic-Guided Test-Time Adaptation for Metric Geometry Estimation” 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
Jiaxing Chen, Zizheng Yan, Jingyu Zhuang, Jinwei Chen, Shuguang Cui, Guanying Chen, Qingnan Fan
IntriAdapt adapts a pretrained geometry model using known camera intrinsics, without depth measurements. It compares predicted point directions with camera-defined pixel rays and updates a lightweight LoRA adapter while keeping the backbone frozen. Across six leakage-free sequence benchmarks, adapted MoGe2 has lower metric point and depth error than its oracle-intrinsics baseline, though some threshold-accuracy scores favor other methods.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.