DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery
We introduce a novel task, Generalized Facial Expression Category Discovery (G-FACE), that discovers new, unseen facial expressions while recognizing known categories effectively. Even though there are generalized category discovery methods for natural images, they show compromised performance on G-...
Na minha lista:
| Autores principales: | , , , , , , , , |
|---|---|
| Formato: | Artigo |
| Publicado em: |
2024
|
| Assuntos: | |
| Acesso em linha: | https://arxiv.org/abs/2409.20098 http://arxiv.org/abs/2409.20098 |
| Tags: |
Sem tags, seja o primeiro a adicionar uma tag!
|
| _version_ | 1879176359739129856 |
|---|---|
| author | Luo, Tingzhang Liu, Yichao Liu, Yuanyuan Zhang, Andi Wang, Xin Zhan, Yibing Tang, Chang Liu, Leyuan Chen, Zhe |
| author_browse | Chen, Zhe Liu, Leyuan Liu, Yichao Liu, Yuanyuan Luo, Tingzhang Tang, Chang Wang, Xin Zhan, Yibing Zhang, Andi |
| author_facet | Luo, Tingzhang Liu, Yichao Liu, Yuanyuan Zhang, Andi Wang, Xin Zhan, Yibing Tang, Chang Liu, Leyuan Chen, Zhe |
| author_sort | Luo, Tingzhang |
| building | arXiv |
| collection | Acceso abierto |
| description | We introduce a novel task, Generalized Facial Expression Category Discovery (G-FACE), that discovers new, unseen facial expressions while recognizing known categories effectively. Even though there are generalized category discovery methods for natural images, they show compromised performance on G-FACE. We identified two biases that affect the learning: implicit bias, coming from an underlying distributional gap between new categories in unlabeled data and known categories in labeled data, and explicit bias, coming from shifted preference on explicit visual facial change characteristics from known expressions to unknown expressions. By addressing the challenges caused by both biases, we propose a Debiased G-FACE method, namely DIG-FACE, that facilitates the debiasing of both implicit and explicit biases. In the implicit debiasing process of DIG-FACE, we devise a novel learning strategy that aims at estimating and minimizing the upper bound of implicit bias. In the explicit debiasing process, we optimize the model's ability to handle nuanced visual facial expression data by introducing a hierarchical category-discrimination refinement strategy: sample-level, triplet-level, and distribution-level optimizations. Extensive experiments demonstrate that our DIG-FACE significantly enhances recognition accuracy for both known and new categories, setting a first-of-its-kind standard for the task. |
| first_indexed | 2026-10-16T00:43:43Z |
| format | Article |
| fullrecord | <oai_dc:dc xmlns:oai_dc="http://www.openarchives.org/OAI/2.0/oai_dc/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ https://www.openarchives.org/OAI/2.0/oai_dc.xsd"><identifier>arxiv-2409.20098</identifier><datestamp>2024-11-20</datestamp>
<dc:title>DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery</dc:title>
<dc:creator>Luo, Tingzhang</dc:creator>
<dc:creator>Liu, Yichao</dc:creator>
<dc:creator>Liu, Yuanyuan</dc:creator>
<dc:creator>Zhang, Andi</dc:creator>
<dc:creator>Wang, Xin</dc:creator>
<dc:creator>Zhan, Yibing</dc:creator>
<dc:creator>Tang, Chang</dc:creator>
<dc:creator>Liu, Leyuan</dc:creator>
<dc:creator>Chen, Zhe</dc:creator>
<dc:subject>Computer Vision and Pattern Recognition</dc:subject>
<dc:description>We introduce a novel task, Generalized Facial Expression Category Discovery (G-FACE), that discovers new, unseen facial expressions while recognizing known categories effectively. Even though there are generalized category discovery methods for natural images, they show compromised performance on G-FACE. We identified two biases that affect the learning: implicit bias, coming from an underlying distributional gap between new categories in unlabeled data and known categories in labeled data, and explicit bias, coming from shifted preference on explicit visual facial change characteristics from known expressions to unknown expressions. By addressing the challenges caused by both biases, we propose a Debiased G-FACE method, namely DIG-FACE, that facilitates the debiasing of both implicit and explicit biases. In the implicit debiasing process of DIG-FACE, we devise a novel learning strategy that aims at estimating and minimizing the upper bound of implicit bias. In the explicit debiasing process, we optimize the model's ability to handle nuanced visual facial expression data by introducing a hierarchical category-discrimination refinement strategy: sample-level, triplet-level, and distribution-level optimizations. Extensive experiments demonstrate that our DIG-FACE significantly enhances recognition accuracy for both known and new categories, setting a first-of-its-kind standard for the task. </dc:description>
<dc:date>2024-09-30</dc:date>
<dc:date>2024-11-19</dc:date>
<dc:type>text</dc:type>
<dc:identifier>https://arxiv.org/abs/2409.20098</dc:identifier>
<dc:identifier>http://arxiv.org/abs/2409.20098</dc:identifier>
</oai_dc:dc> |
| id | arxiv-2409.20098 |
| institution | arXiv |
| last_indexed | 2026-10-16T00:43:43Z |
| publishDate | 2024 |
| publishDateRange | 2024 |
| publishDateSort | 2024 |
| record_format | oai_dc |
| spelling | arxiv-2409.200982024-11-20 DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery Luo, Tingzhang Liu, Yichao Liu, Yuanyuan Zhang, Andi Wang, Xin Zhan, Yibing Tang, Chang Liu, Leyuan Chen, Zhe Computer Vision and Pattern Recognition We introduce a novel task, Generalized Facial Expression Category Discovery (G-FACE), that discovers new, unseen facial expressions while recognizing known categories effectively. Even though there are generalized category discovery methods for natural images, they show compromised performance on G-FACE. We identified two biases that affect the learning: implicit bias, coming from an underlying distributional gap between new categories in unlabeled data and known categories in labeled data, and explicit bias, coming from shifted preference on explicit visual facial change characteristics from known expressions to unknown expressions. By addressing the challenges caused by both biases, we propose a Debiased G-FACE method, namely DIG-FACE, that facilitates the debiasing of both implicit and explicit biases. In the implicit debiasing process of DIG-FACE, we devise a novel learning strategy that aims at estimating and minimizing the upper bound of implicit bias. In the explicit debiasing process, we optimize the model's ability to handle nuanced visual facial expression data by introducing a hierarchical category-discrimination refinement strategy: sample-level, triplet-level, and distribution-level optimizations. Extensive experiments demonstrate that our DIG-FACE significantly enhances recognition accuracy for both known and new categories, setting a first-of-its-kind standard for the task. 2024-09-30 2024-11-19 text https://arxiv.org/abs/2409.20098 http://arxiv.org/abs/2409.20098 |
| spellingShingle | Computer Vision and Pattern Recognition Luo, Tingzhang Liu, Yichao Liu, Yuanyuan Zhang, Andi Wang, Xin Zhan, Yibing Tang, Chang Liu, Leyuan Chen, Zhe DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery |
| title | DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery |
| title_full | DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery |
| title_fullStr | DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery |
| title_full_unstemmed | DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery |
| title_short | DIG-FACE: De-biased Learning for Generalized Facial Expression Category Discovery |
| title_sort | dig face de biased learning for generalized facial expression category discovery |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2409.20098 http://arxiv.org/abs/2409.20098 |
| work_keys_str_mv | AT luotingzhang digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT liuyichao digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT liuyuanyuan digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT zhangandi digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT wangxin digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT zhanyibing digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT tangchang digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT liuleyuan digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery AT chenzhe digfacedebiasedlearningforgeneralizedfacialexpressioncategorydiscovery |