Reproduce “Decoupled PROB: Decoupled Query Initialization Tasks and Objectness-Class Learning 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
Riku Inoue, Masamitsu Tsuchiya, Yuji Yasui
Decoupled PROB addresses open-world object detection by assigning queries separate roles for known-object search and unknown-object coverage. It combines encoder-selected queries with broadly distributed learnable queries, then stops objectness prediction after decoder layer 2 while class and box refinement continue through all six layers. On M-OWODB Task 2, current-known [email protected] is 36.7 versus PROB’s 32.2, but gains are not uniform.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.