Reproduce “Test-Time Adaptive Object Detection with Foundation 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
Yingjie Gao, Yanan Zhang, Zhi Cai, Di Huang
The authors present MPMT, which adapts pretrained Grounding DINO to unlabeled test images by tuning text and image prompts and reusing high-confidence object crops from earlier images. On Pascal-C, it reaches 56.2% average AP50 across 15 corruption types, versus 51.5% for matched Mean-Teacher. Across 13 datasets, it averages 54.2% mAP, with exceptions on two very small targets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.