Deep multi-negative supervised hashing for large-scale image retrieval

Yingfan Liu, Xiaotian Qiao, Zhaoqing Liu, Xiaofang Xia, Yinlong Zhang, Jiangtao Cui

Abstract: Highlights•Sampling multiple negatives to leverage more structure and supervise information.•Using a limited number of training images to construct multi-negative tuples.•The multi-negative loss function preserves the similarity semantics of images.