PLDH: Pseudo-Labels Based Deep Hashing
Deep hashing has received a great deal of attraction in large-scale data analysis, due to its high efficiency and effectiveness. The performance of deep hashing models heavily relies on label information, which is very expensive to obtain. In this work, a novel end-to-end deep hashing model based on...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/11/9/2175 |
| _version_ | 1797602176686620672 |
|---|---|
| author | Huawen Liu Minhao Yin Zongda Wu Liping Zhao Qi Li Xinzhong Zhu Zhonglong Zheng |
| author_facet | Huawen Liu Minhao Yin Zongda Wu Liping Zhao Qi Li Xinzhong Zhu Zhonglong Zheng |
| author_sort | Huawen Liu |
| collection | DOAJ |
| description | Deep hashing has received a great deal of attraction in large-scale data analysis, due to its high efficiency and effectiveness. The performance of deep hashing models heavily relies on label information, which is very expensive to obtain. In this work, a novel end-to-end deep hashing model based on pseudo-labels for large-scale data without labels is proposed. The proposed hashing model consists of two major stages, where the first stage aims to obtain pseudo-labels based on deep features extracted by a pre-training deep convolution neural network. The second stage generates hash codes with high quality by the same neural network in the previous stage, coupled with an end-to-end hash layer, whose purpose is to encode data into a binary representation. Additionally, a quantization loss is introduced and interwound within these two stages. Evaluation experiments were conducted on two frequently-used image collections, CIFAR-10 and NUS-WIDE, with eight popular shallow and deep hashing models. The experimental results show the superiority of the proposed method in image retrieval. |
| first_indexed | 2024-03-11T04:13:22Z |
| format | Article |
| id | doaj.art-98ee9b36931b4dfdb488cea9ff3b4a61 |
| institution | Directory Open Access Journal |
| issn | 2227-7390 |
| language | English |
| last_indexed | 2024-03-11T04:13:22Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj.art-98ee9b36931b4dfdb488cea9ff3b4a612023-11-17T23:20:58ZengMDPI AGMathematics2227-73902023-05-01119217510.3390/math11092175PLDH: Pseudo-Labels Based Deep HashingHuawen Liu0Minhao Yin1Zongda Wu2Liping Zhao3Qi Li4Xinzhong Zhu5Zhonglong Zheng6Department of Computer Science, Shaoxing University, Shaoxing 312000, ChinaSchool of Information Science and Technology, Northeast Normal University, Changchun 130024, ChinaDepartment of Computer Science, Shaoxing University, Shaoxing 312000, ChinaDepartment of Computer Science, Shaoxing University, Shaoxing 312000, ChinaDepartment of Computer Science, Shaoxing University, Shaoxing 312000, ChinaSchool of Computer Science and Technology, Zhejiang Normal University, Jinhua 311231, ChinaSchool of Computer Science and Technology, Zhejiang Normal University, Jinhua 311231, ChinaDeep hashing has received a great deal of attraction in large-scale data analysis, due to its high efficiency and effectiveness. The performance of deep hashing models heavily relies on label information, which is very expensive to obtain. In this work, a novel end-to-end deep hashing model based on pseudo-labels for large-scale data without labels is proposed. The proposed hashing model consists of two major stages, where the first stage aims to obtain pseudo-labels based on deep features extracted by a pre-training deep convolution neural network. The second stage generates hash codes with high quality by the same neural network in the previous stage, coupled with an end-to-end hash layer, whose purpose is to encode data into a binary representation. Additionally, a quantization loss is introduced and interwound within these two stages. Evaluation experiments were conducted on two frequently-used image collections, CIFAR-10 and NUS-WIDE, with eight popular shallow and deep hashing models. The experimental results show the superiority of the proposed method in image retrieval.https://www.mdpi.com/2227-7390/11/9/2175learning to hashimage retrievaldeep learningnearest neighbor searchunsupervised learningpseudo-label |
| spellingShingle | Huawen Liu Minhao Yin Zongda Wu Liping Zhao Qi Li Xinzhong Zhu Zhonglong Zheng PLDH: Pseudo-Labels Based Deep Hashing Mathematics learning to hash image retrieval deep learning nearest neighbor search unsupervised learning pseudo-label |
| title | PLDH: Pseudo-Labels Based Deep Hashing |
| title_full | PLDH: Pseudo-Labels Based Deep Hashing |
| title_fullStr | PLDH: Pseudo-Labels Based Deep Hashing |
| title_full_unstemmed | PLDH: Pseudo-Labels Based Deep Hashing |
| title_short | PLDH: Pseudo-Labels Based Deep Hashing |
| title_sort | pldh pseudo labels based deep hashing |
| topic | learning to hash image retrieval deep learning nearest neighbor search unsupervised learning pseudo-label |
| url | https://www.mdpi.com/2227-7390/11/9/2175 |
| work_keys_str_mv | AT huawenliu pldhpseudolabelsbaseddeephashing AT minhaoyin pldhpseudolabelsbaseddeephashing AT zongdawu pldhpseudolabelsbaseddeephashing AT lipingzhao pldhpseudolabelsbaseddeephashing AT qili pldhpseudolabelsbaseddeephashing AT xinzhongzhu pldhpseudolabelsbaseddeephashing AT zhonglongzheng pldhpseudolabelsbaseddeephashing |