Pyramid hybrid pooling quantization for efficient fine-grained image retrieval

Ziyun Zeng, Jinpeng Wang, Bin Chen, Tao Dai, Shu-Tao Xia, Zhi Wang

Abstract: Highlights•We propose Pyramid Hybrid Pooling Quantization for fine-grained image retrieval and achieve new state-of-the-art.•A Pyramid Hybrid Pooling model is proposed to preserve fine-grained information, yielding discriminative representations.•A partial codebook attention mechanism is proposed to better quantize features by attentively optimizing relevant codewords.