Reproduce “RbA: Segmenting Unknown Regions Rejected by All” 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
Nazir Nayal, Mısra Yavuz, João F. Henriques, Fatma Güney
RbA identifies unknown road-scene regions by asking whether known classes collectively reject them, rather than treating any uncertain pixel as anomalous. It derives an outlier score from mask-classification region predictions and supports outlier training. On Road Anomaly, Swin-B reaches 85.42 pixel-level AP and 6.92 FPR@95, while retaining 82.20 mIoU; these benchmark results do not establish live-driving safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.