2021/05/28 by Han Zhang, Hongwei Shen, Yiming Qiu +6 · 1 citation
Computer Science · #Advanced Image and Video Retrieval Techniques #Deep learning #Domain Adaptation and Few-Shot Learning #Embedding #Image Retrieval and Classification Techniques #Outer product #Pattern recognition (psychology) #Product (mathematics) #Quantization (signal processing) #Search engine indexing #cs.IR
paper · pdf · doi:10.1145/3404835.3462988
4 pages, 4 figures; accepted by SIGIR2021
openalex created_date 2021/04/26 · arxiv created 2021/05/28 · arxiv updated 2021/05/31 · openalex publication_date 2021/07/11 · openalex updated_date 2026/08/05
Embedding index that enables fast approximate nearest neighbor(ANN) search, serves as an indispensable component for state-of-the-art deep retrieval systems. Traditional approaches, often separating the two steps of embedding learning and index building, incur additional indexing time and decayed retrieval accuracy. In this paper, we propose a novel method called Poeem, which stands for product quantization based embedding index jointly trained with deep retrieval model, to unify the two separate steps within an end-to-end training, by utilizing a few techniques including the gradient straight-through estimator, warm start strategy, optimal space decomposition and Givens rotation. Extensive experimental results show that the proposed method not only improves retrieval accuracy significantly but also reduces the indexing time to almost none. We have open sourced our approach for the sake of comparison and reproducibility.