English

Efficient Cross-Modal Retrieval via Deep Binary Hashing and Quantization

Information Retrieval 2022-02-22 v1 Computer Vision and Pattern Recognition Machine Learning

Abstract

Cross-modal retrieval aims to search for data with similar semantic meanings across different content modalities. However, cross-modal retrieval requires huge amounts of storage and retrieval time since it needs to process data in multiple modalities. Existing works focused on learning single-source compact features such as binary hash codes that preserve similarities between different modalities. In this work, we propose a jointly learned deep hashing and quantization network (HQ) for cross-modal retrieval. We simultaneously learn binary hash codes and quantization codes to preserve semantic information in multiple modalities by an end-to-end deep learning architecture. At the retrieval step, binary hashing is used to retrieve a subset of items from the search space, then quantization is used to re-rank the retrieved items. We theoretically and empirically show that this two-stage retrieval approach provides faster retrieval results while preserving accuracy. Experimental results on the NUS-WIDE, MIR-Flickr, and Amazon datasets demonstrate that HQ achieves boosts of more than 7% in precision compared to supervised neural network-based compact coding models.

Keywords

Cite

@article{arxiv.2202.10232,
  title  = {Efficient Cross-Modal Retrieval via Deep Binary Hashing and Quantization},
  author = {Yang Shi and Young-joo Chung},
  journal= {arXiv preprint arXiv:2202.10232},
  year   = {2022}
}

Comments

Accepted at BMVC 2021

R2 v1 2026-06-24T09:47:48.565Z