Reproduce “RoSAMDepth: Robust Self-supervised Depth Estimation Leveraging Segment Anything Model” 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
Xuanang Gao, Zhiwei Ning, Gengming Zhang, Jiaxi Cao, Runze Yang, Zhonglong Zheng, Jie Yang, Rong Xiao and 1 more
RoSAMDepth uses pre-generated SAM regions to guide self-supervised monocular depth training. It shapes decoder features, identifies unreliable teacher-depth outliers within regions, and downweights unreliable region-level supervision. On single-frame Oxford RobotCar nighttime evaluation over 0.1–50 m, it reports AbsRel of 0.1066 versus 0.1103 for Syn2Real-Depth. Its reported gains cover almost all metrics, not every setting; multi-frame nuScenes nighttime RMSE is worse.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.