Discriminative dual-stream deep hashing for large-scale image retrieval

Yujuan Ding, Wai Keung Wong, Zhihui Lai, Zheng Zhang

Abstract

Highlights • Most existing deep supervised hashing methods mainly focus on how to effectively preserve the similarity information of semantic labels while ignoring discrimination information in the labels. • A novel discriminative dual-stream deep hashing method is proposed to integrate the pairwise similarity loss and the classification loss into a unified framework to take full advantage of label information. • The proposed method enlarges the margin between the different classes so that can generate discrimination of learned binary codes for better image retrieval performance. • Extensive experiments show that the proposed method consistently outperforms current state-of- the-art methods on three benchmark datasets for image retrieval task. Abstract Deep hashing has been an important research topic for using deep learning to boost performance of hash learning. Most existing deep supervised hashing methods mainly focus on how to effectively preserve the similarity in hash coding solely depending on pairwise supervision. However, such pairwise similarity-preserving strategy cannot fully explore the semantic information in most cases, which results in information loss. To address this problem, this paper proposes a discriminative dual-stream deep hashing (DDDH) method, which integrates the pairwise similarity loss and the classification loss into a unified framework to take full advantage of label information. Specifically, the pairwise similarity loss aims to preserve the similarity and structural information of high-dimensional original data. Meanwhile, the designed classification loss can enlarge the margin between different classes which improves the discrimination of learned binary codes. Moreover, an effective optimization algorithm is employed to train the hash code learning framework in an end-to-end manner. The results of extensive experiments on three image datasets demonstrate that our method is superior to several state-of-the-art deep and non-deep hashing methods. Ablation studies and analysis further show the effectiveness of introducing the classification loss in the overall hash learning framework.
BibTeX
@article{ding2020discriminative,
  author = {Yujuan Ding and Wai Keung Wong and Zhihui Lai and Zheng Zhang},
  title = {{Discriminative dual-stream deep hashing for large-scale image retrieval}},
  journal = {Information Processing \& Management},
  year = {2020},
  doi = {10.1016/j.ipm.2020.102288},
}

Loading related content...

Figure