Reproduce “Learning Multi-Modal Prototypes for Cross-Domain Few-Shot 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
Wanqi Wang, Jingcai Guo, Yuxiang Cai, Zhi Chen
LMP supplements text-guided detection with visual references built from labeled target examples and hard-negative regions drawn from confusing query-image areas during training. The authors fine-tune pretrained GroundingDINO on few-shot target data without source-domain training data. In the 1-shot setting, LMP reaches 34.3 average mAP across six domains, versus 33.6 for Domain-RAG, but trails it on two domains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.