Skip to content
Back to outputs

Cosine metric supervised deep hashing with balanced similarity

Research output: Contribution to journalArticlepeer-review

Standard

Cosine metric supervised deep hashing with balanced similarity. / Hu, Wenjin; Wu, Lifang; Jian, Meng; Chen, Yukun; Yu, Hui.

In: Neurocomputing, 25.03.2021.

Research output: Contribution to journalArticlepeer-review

Harvard

Hu, W, Wu, L, Jian, M, Chen, Y & Yu, H 2021, 'Cosine metric supervised deep hashing with balanced similarity', Neurocomputing.

APA

Hu, W., Wu, L., Jian, M., Chen, Y., & Yu, H. (Accepted/In press). Cosine metric supervised deep hashing with balanced similarity. Neurocomputing.

Vancouver

Hu W, Wu L, Jian M, Chen Y, Yu H. Cosine metric supervised deep hashing with balanced similarity. Neurocomputing. 2021 Mar 25.

Author

Hu, Wenjin ; Wu, Lifang ; Jian, Meng ; Chen, Yukun ; Yu, Hui. / Cosine metric supervised deep hashing with balanced similarity. In: Neurocomputing. 2021.

Bibtex

@article{46c49d1351d54ab5abe582d1d22c536c,
title = "Cosine metric supervised deep hashing with balanced similarity",
abstract = "Deep supervised hashing takes prominent advantages of low storage cost, high computational efficiency and good retrieval performance, which draws attention in the field of large-scale image retrieval. However, similarity-preserving, quantization errors and imbalanced data are still great challenges in deep supervised hashing. This paper proposes a pairwise similarity-preserving deep hashing scheme to handle the aforementioned problems in a unified framework, termed as Cosine Metric Supervised Deep Hashing with Balanced Similarity (BCMDH). BCMDH integrates contrastive cosine similarity and Cosine distance entropy quantization to preserve the original semantic distribution and reduce the quantization errors simultaneously. Furthermore, a weighted similarity measure with cosine metric entropy is designed to reduce the impact of imbalanced data, which adaptively assigns weights according to sample attributes (pos/neg and easy/hard) in the embedding process of similarity-preserving. The experimental results on four widely-used datasets demonstrate that the proposed method is capable of generating hash codes of high quality and improve largescale image retrieval performance.",
author = "Wenjin Hu and Lifang Wu and Meng Jian and Yukun Chen and Hui Yu",
note = "12 months embargo. Elsevier.",
year = "2021",
month = mar,
day = "25",
language = "English",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",

}

RIS

TY - JOUR

T1 - Cosine metric supervised deep hashing with balanced similarity

AU - Hu, Wenjin

AU - Wu, Lifang

AU - Jian, Meng

AU - Chen, Yukun

AU - Yu, Hui

N1 - 12 months embargo. Elsevier.

PY - 2021/3/25

Y1 - 2021/3/25

N2 - Deep supervised hashing takes prominent advantages of low storage cost, high computational efficiency and good retrieval performance, which draws attention in the field of large-scale image retrieval. However, similarity-preserving, quantization errors and imbalanced data are still great challenges in deep supervised hashing. This paper proposes a pairwise similarity-preserving deep hashing scheme to handle the aforementioned problems in a unified framework, termed as Cosine Metric Supervised Deep Hashing with Balanced Similarity (BCMDH). BCMDH integrates contrastive cosine similarity and Cosine distance entropy quantization to preserve the original semantic distribution and reduce the quantization errors simultaneously. Furthermore, a weighted similarity measure with cosine metric entropy is designed to reduce the impact of imbalanced data, which adaptively assigns weights according to sample attributes (pos/neg and easy/hard) in the embedding process of similarity-preserving. The experimental results on four widely-used datasets demonstrate that the proposed method is capable of generating hash codes of high quality and improve largescale image retrieval performance.

AB - Deep supervised hashing takes prominent advantages of low storage cost, high computational efficiency and good retrieval performance, which draws attention in the field of large-scale image retrieval. However, similarity-preserving, quantization errors and imbalanced data are still great challenges in deep supervised hashing. This paper proposes a pairwise similarity-preserving deep hashing scheme to handle the aforementioned problems in a unified framework, termed as Cosine Metric Supervised Deep Hashing with Balanced Similarity (BCMDH). BCMDH integrates contrastive cosine similarity and Cosine distance entropy quantization to preserve the original semantic distribution and reduce the quantization errors simultaneously. Furthermore, a weighted similarity measure with cosine metric entropy is designed to reduce the impact of imbalanced data, which adaptively assigns weights according to sample attributes (pos/neg and easy/hard) in the embedding process of similarity-preserving. The experimental results on four widely-used datasets demonstrate that the proposed method is capable of generating hash codes of high quality and improve largescale image retrieval performance.

M3 - Article

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

ER -

ID: 26920281