Reproduce “Unsupervised Data-Efficient Cross-Modal Retrieval with Global-Neighborhood Alignment Hashing” 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
Runhao Li, Xiaoxu Ma, Zhenyu Weng, Yue Zhang, Guibo Luo, Huiping Zhuang, Zhiping Lin, Yap-Peng Tan
GNAH enables unsupervised cross-modal retrieval by mapping image and text features into shared binary codes. It transfers global structure from frozen CLIP representations through prototypes and aligns sampled local neighborhoods. On MIR Flickr, with 80 training pairs and 32-bit codes, it reports mAP of 0.686 for I2T and 0.685 for T2I, benchmark results under the sampled-pair protocol.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.