Reproduce “Beyond Proposals: Region-Consistent Open-Set 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
Jiehui Tang, Yunlong Yu, Zhao Wang, Jungong Han
OCSI-Net reconciles known and unknown predictions from overlapping boxes by using shared region-level evidence and a geometry-aware training method. On four benchmarks under OSOD-III, its mean H, the harmonic mean of known and unknown average precision, was 34.12, versus OpenDet’s 27.29, a difference of 6.83 points. It reports the highest unknown average precision on three of four benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.