Reproduce “PROB: Probabilistic Objectness for Open World Object Detection” 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
Orr Zohar, Kuan-Chieh Wang, Serena Yeung
PROB separates objectness from known-class prediction in open-world object detection. It learns a class-agnostic probabilistic objectness score from query embeddings and uses that score to gate class predictions. On M-OWODB Task 1, PROB achieves unknown-object recall of 19.4 versus 7.5 for OW-DETR, with known-class [email protected] of 59.5 versus 59.2; these are benchmark results, not evidence of deployment reliability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.