Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing
In recent years, unsupervised deep hashing methods have achieved great success in large-scale image retrieval. However, these approaches still suffer two major problems in real world applications. On the one hand, due to the lack of effective supervision information, hash codes of different categori...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2020-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9039610/ |
| _version_ | 1828406569938714624 |
|---|---|
| author | Huiying Li Yang Li Xin Xie Shuai Gao Dongsheng Mao |
| author_facet | Huiying Li Yang Li Xin Xie Shuai Gao Dongsheng Mao |
| author_sort | Huiying Li |
| collection | DOAJ |
| description | In recent years, unsupervised deep hashing methods have achieved great success in large-scale image retrieval. However, these approaches still suffer two major problems in real world applications. On the one hand, due to the lack of effective supervision information, hash codes of different categories are easily judged to be similar. On the other hand, binary semantic similarity matrices can not reflect ranking relationship and the internal structure information of different images. To solve these problems, we propose a novel unsupervised deep hashing method, named <bold>P</bold>seudo labels and <bold>S</bold>oft multi-part <bold>C</bold>orresponding similarity based <bold>H</bold>ashing (PSCH), to ensure the heterogeneity of the hash codes. Specifically, we propose a “pseudo labels” method that use <inline-formula> <tex-math notation="LaTeX">${k}$ </tex-math></inline-formula>-means clustering and a distance threshold to generate the pseudo labels. Further, in order to reflect the hash codes similarity between instances within the same class, we propose a novel soft multi-part corresponding similarity method to learn better hash codes. This method can divide deep feature maps into several groups and compute the attention map for multi-part similarity matrices. In addition, a novel loss function is proposed to support learning with pseudo labels and soft multi-part corresponding similarity for achieving better performance. Comprehensive experiments on CIFAR-10, NUSWIDE, and Flickr demonstrate that our method can generate high-quality hash codes and outperform state-of-the-art unsupervised hashing methods by a large margin. |
| first_indexed | 2024-12-10T11:13:47Z |
| format | Article |
| id | doaj.art-a234e9b0217640499cfb20916a72cfcd |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-12-10T11:13:47Z |
| publishDate | 2020-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-a234e9b0217640499cfb20916a72cfcd2022-12-22T01:51:18ZengIEEEIEEE Access2169-35362020-01-018535115352110.1109/ACCESS.2020.29812889039610Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep HashingHuiying Li0https://orcid.org/0000-0002-8123-7768Yang Li1https://orcid.org/0000-0003-1682-0284Xin Xie2https://orcid.org/0000-0002-4398-5407Shuai Gao3https://orcid.org/0000-0001-6802-9553Dongsheng Mao4https://orcid.org/0000-0002-4855-6492Command and Control Engineering College, Army Engineering University of PLA, Nanjing, ChinaCommand and Control Engineering College, Army Engineering University of PLA, Nanjing, ChinaCommand and Control Engineering College, Army Engineering University of PLA, Nanjing, ChinaCommand and Control Engineering College, Army Engineering University of PLA, Nanjing, ChinaCommand and Control Engineering College, Army Engineering University of PLA, Nanjing, ChinaIn recent years, unsupervised deep hashing methods have achieved great success in large-scale image retrieval. However, these approaches still suffer two major problems in real world applications. On the one hand, due to the lack of effective supervision information, hash codes of different categories are easily judged to be similar. On the other hand, binary semantic similarity matrices can not reflect ranking relationship and the internal structure information of different images. To solve these problems, we propose a novel unsupervised deep hashing method, named <bold>P</bold>seudo labels and <bold>S</bold>oft multi-part <bold>C</bold>orresponding similarity based <bold>H</bold>ashing (PSCH), to ensure the heterogeneity of the hash codes. Specifically, we propose a “pseudo labels” method that use <inline-formula> <tex-math notation="LaTeX">${k}$ </tex-math></inline-formula>-means clustering and a distance threshold to generate the pseudo labels. Further, in order to reflect the hash codes similarity between instances within the same class, we propose a novel soft multi-part corresponding similarity method to learn better hash codes. This method can divide deep feature maps into several groups and compute the attention map for multi-part similarity matrices. In addition, a novel loss function is proposed to support learning with pseudo labels and soft multi-part corresponding similarity for achieving better performance. Comprehensive experiments on CIFAR-10, NUSWIDE, and Flickr demonstrate that our method can generate high-quality hash codes and outperform state-of-the-art unsupervised hashing methods by a large margin.https://ieeexplore.ieee.org/document/9039610/Unsupervised hashingpseudo labelsimage retrievalmulti-part correspondencesoft similarity |
| spellingShingle | Huiying Li Yang Li Xin Xie Shuai Gao Dongsheng Mao Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing IEEE Access Unsupervised hashing pseudo labels image retrieval multi-part correspondence soft similarity |
| title | Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing |
| title_full | Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing |
| title_fullStr | Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing |
| title_full_unstemmed | Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing |
| title_short | Pseudo Labels and Soft Multi-Part Corresponding Similarity for Unsupervised Deep Hashing |
| title_sort | pseudo labels and soft multi part corresponding similarity for unsupervised deep hashing |
| topic | Unsupervised hashing pseudo labels image retrieval multi-part correspondence soft similarity |
| url | https://ieeexplore.ieee.org/document/9039610/ |
| work_keys_str_mv | AT huiyingli pseudolabelsandsoftmultipartcorrespondingsimilarityforunsuperviseddeephashing AT yangli pseudolabelsandsoftmultipartcorrespondingsimilarityforunsuperviseddeephashing AT xinxie pseudolabelsandsoftmultipartcorrespondingsimilarityforunsuperviseddeephashing AT shuaigao pseudolabelsandsoftmultipartcorrespondingsimilarityforunsuperviseddeephashing AT dongshengmao pseudolabelsandsoftmultipartcorrespondingsimilarityforunsuperviseddeephashing |