Beyond Product Quantization: Deep Progressive Quantization for Image Retrieval
This addresses the problem of efficient and flexible quantization for large-scale image retrieval, offering a novel method that reduces retraining needs and improves performance, though it is incremental in the context of quantization techniques.
The paper tackles the limitations of Product Quantization in image retrieval by proposing Deep Progressive Quantization, which learns quantization codes sequentially to approximate features progressively, enabling training once for different code lengths and significantly outperforming state-of-the-art methods on benchmark datasets.
Product Quantization (PQ) has long been a mainstream for generating an exponentially large codebook at very low memory/time cost. Despite its success, PQ is still tricky for the decomposition of high-dimensional vector space, and the retraining of model is usually unavoidable when the code length changes. In this work, we propose a deep progressive quantization (DPQ) model, as an alternative to PQ, for large scale image retrieval. DPQ learns the quantization codes sequentially and approximates the original feature space progressively. Therefore, we can train the quantization codes with different code lengths simultaneously. Specifically, we first utilize the label information for guiding the learning of visual features, and then apply several quantization blocks to progressively approach the visual features. Each quantization block is designed to be a layer of a convolutional neural network, and the whole framework can be trained in an end-to-end manner. Experimental results on the benchmark datasets show that our model significantly outperforms the state-of-the-art for image retrieval. Our model is trained once for different code lengths and therefore requires less computation time. Additional ablation study demonstrates the effect of each component of our proposed model. Our code is released at https://github.com/cfm-uestc/DPQ.