Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation
Precise iris segmentation is a very important part of accurate iris recognition. Traditional iris segmentation methods require complex prior knowledge and pre- and post-processing and have limited accuracy under non-ideal conditions. Deep learning approaches outperform traditional methods. However,...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-09-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/24/9/1276 |
| _version_ | 1827661232877862912 |
|---|---|
| author | Zhiyong Zhou Yuanning Liu Xiaodong Zhu Shuai Liu Shaoqiang Zhang Yuanfeng Li |
| author_facet | Zhiyong Zhou Yuanning Liu Xiaodong Zhu Shuai Liu Shaoqiang Zhang Yuanfeng Li |
| author_sort | Zhiyong Zhou |
| collection | DOAJ |
| description | Precise iris segmentation is a very important part of accurate iris recognition. Traditional iris segmentation methods require complex prior knowledge and pre- and post-processing and have limited accuracy under non-ideal conditions. Deep learning approaches outperform traditional methods. However, the limitation of a small number of labeled datasets degrades their performance drastically because of the difficulty in collecting and labeling irises. Furthermore, previous approaches ignore the large distribution gap within the non-ideal iris dataset due to illumination, motion blur, squinting eyes, etc. To address these issues, we propose a three-stage training strategy. Firstly, supervised contrastive pretraining is proposed to increase intra-class compactness and inter-class separability to obtain a good pixel classifier under a limited amount of data. Secondly, the entire network is fine-tuned using cross-entropy loss. Thirdly, an intra-dataset adversarial adaptation is proposed, which reduces the intra-dataset gap in the non-ideal situation by aligning the distribution of the hard and easy samples at the pixel class level. Our experiments show that our method improved the segmentation performance and achieved the following encouraging results: 0.44%, 1.03%, 0.66%, 0.41%, and 0.37% in the <i>Nice1</i> and 96.66%, 98.72%, 93.21%, 94.28%, and 97.41% in the <i>F1</i> for UBIRIS.V2, IITD, MICHE-I, CASIA-D, and CASIA-T. |
| first_indexed | 2024-03-10T00:05:35Z |
| format | Article |
| id | doaj.art-b5f8d0a93842454c885f31e5d4c276db |
| institution | Directory Open Access Journal |
| issn | 1099-4300 |
| language | English |
| last_indexed | 2024-03-10T00:05:35Z |
| publishDate | 2022-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj.art-b5f8d0a93842454c885f31e5d4c276db2023-11-23T16:08:56ZengMDPI AGEntropy1099-43002022-09-01249127610.3390/e24091276Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris SegmentationZhiyong Zhou0Yuanning Liu1Xiaodong Zhu2Shuai Liu3Shaoqiang Zhang4Yuanfeng Li5College of Computer Science and Technology, Jilin University, Changchun 130012, ChinaCollege of Computer Science and Technology, Jilin University, Changchun 130012, ChinaCollege of Computer Science and Technology, Jilin University, Changchun 130012, ChinaCollege of Computer Science and Technology, Jilin University, Changchun 130012, ChinaCollege of Computer Science and Technology, Jilin University, Changchun 130012, ChinaCollege of Biological and Agricultural Engineering, Jilin University, Changchun 130012, ChinaPrecise iris segmentation is a very important part of accurate iris recognition. Traditional iris segmentation methods require complex prior knowledge and pre- and post-processing and have limited accuracy under non-ideal conditions. Deep learning approaches outperform traditional methods. However, the limitation of a small number of labeled datasets degrades their performance drastically because of the difficulty in collecting and labeling irises. Furthermore, previous approaches ignore the large distribution gap within the non-ideal iris dataset due to illumination, motion blur, squinting eyes, etc. To address these issues, we propose a three-stage training strategy. Firstly, supervised contrastive pretraining is proposed to increase intra-class compactness and inter-class separability to obtain a good pixel classifier under a limited amount of data. Secondly, the entire network is fine-tuned using cross-entropy loss. Thirdly, an intra-dataset adversarial adaptation is proposed, which reduces the intra-dataset gap in the non-ideal situation by aligning the distribution of the hard and easy samples at the pixel class level. Our experiments show that our method improved the segmentation performance and achieved the following encouraging results: 0.44%, 1.03%, 0.66%, 0.41%, and 0.37% in the <i>Nice1</i> and 96.66%, 98.72%, 93.21%, 94.28%, and 97.41% in the <i>F1</i> for UBIRIS.V2, IITD, MICHE-I, CASIA-D, and CASIA-T.https://www.mdpi.com/1099-4300/24/9/1276contrastive learningadversarial adaptationiris segmentationdeep learning |
| spellingShingle | Zhiyong Zhou Yuanning Liu Xiaodong Zhu Shuai Liu Shaoqiang Zhang Yuanfeng Li Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation Entropy contrastive learning adversarial adaptation iris segmentation deep learning |
| title | Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation |
| title_full | Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation |
| title_fullStr | Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation |
| title_full_unstemmed | Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation |
| title_short | Supervised Contrastive Learning and Intra-Dataset Adversarial Adaptation for Iris Segmentation |
| title_sort | supervised contrastive learning and intra dataset adversarial adaptation for iris segmentation |
| topic | contrastive learning adversarial adaptation iris segmentation deep learning |
| url | https://www.mdpi.com/1099-4300/24/9/1276 |
| work_keys_str_mv | AT zhiyongzhou supervisedcontrastivelearningandintradatasetadversarialadaptationforirissegmentation AT yuanningliu supervisedcontrastivelearningandintradatasetadversarialadaptationforirissegmentation AT xiaodongzhu supervisedcontrastivelearningandintradatasetadversarialadaptationforirissegmentation AT shuailiu supervisedcontrastivelearningandintradatasetadversarialadaptationforirissegmentation AT shaoqiangzhang supervisedcontrastivelearningandintradatasetadversarialadaptationforirissegmentation AT yuanfengli supervisedcontrastivelearningandintradatasetadversarialadaptationforirissegmentation |