Reproduce “FACTOR: Counterfactual Training-Free Test-Time Adaptation for Open-Vocabulary 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
Kaixiang Zhao, Mao Ye, Lihua Zhou, Hu Wang, Luping Ji, Song Tang, Xiatian Zhu
FACTOR adapts open-vocabulary detectors without retraining or changing parameters by comparing predictions for each image with those for a version altered by six transformations. It uses region-level sensitivity, visual-attribute alignment, and text-based affinity to recalibrate predictions. On COCO-C with Grounding DINO Swin-T, it reaches 29.30 mAP@50 versus BCA+’s 25.06 across 15 corruption types, on corrupted-image benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.