Rethinking Weak-to-Strong Augmentation in Source-Free Domain Adaptive Object Detection
| Title: | Rethinking Weak-to-Strong Augmentation in Source-Free Domain Adaptive Object Detection |
|---|---|
| Authors: | Yang, Jiuzheng, Tang, Song, Zhang, Yangkuiyi, Li, Shuaifeng, Ye, Mao, Zhang, Jianwei, Zhu, Xiatian |
| Publication Year: | 2024 |
| Collection: | Computer Science |
| Subject Terms: | Computer Science - Computer Vision and Pattern Recognition |
| More Details: | Source-Free domain adaptive Object Detection (SFOD) aims to transfer a detector (pre-trained on source domain) to new unlabelled target domains. Current SFOD methods typically follow the Mean Teacher framework, where weak-to-strong augmentation provides diverse and sharp contrast for self-supervised learning. However, this augmentation strategy suffers from an inherent problem called crucial semantics loss: Due to random, strong disturbance, strong augmentation is prone to losing typical visual components, hindering cross-domain feature extraction. To address this thus-far ignored limitation, this paper introduces a novel Weak-to-Strong Contrastive Learning (WSCoL) approach. The core idea is to distill semantics lossless knowledge in the weak features (from the weak/teacher branch) to guide the representation learning upon the strong features (from the strong/student branch). To achieve this, we project the original features into a shared space using a mapping network, thereby reducing the bias between the weak and strong features. Meanwhile, a weak features-guided contrastive learning is performed in a weak-to-strong manner alternatively. Specifically, we first conduct an adaptation-aware prototype-guided clustering on the weak features to generate pseudo labels for corresponding strong features matched through proposals. Sequentially, we identify positive-negative samples based on the pseudo labels and perform cross-category contrastive learning on the strong features where an uncertainty estimator encourages adaptive background contrast. Extensive experiments demonstrate that WSCoL yields new state-of-the-art performance, offering a built-in mechanism mitigating crucial semantics loss for traditional Mean Teacher framework. The code and data will be released soon. |
| Document Type: | Working Paper |
| Access URL: | http://arxiv.org/abs/2410.05557 |
| Accession Number: | edsarx.2410.05557 |
| Database: | arXiv |
| FullText | Text: Availability: 0 CustomLinks: – Url: http://arxiv.org/abs/2410.05557 Name: EDS - Arxiv Category: fullText Text: View this record from Arxiv MouseOverText: View this record from Arxiv – Url: https://resolver.ebsco.com/c/xy5jbn/result?sid=EBSCO:edsarx&genre=article&issn=&ISBN=&volume=&issue=&date=20241007&spage=&pages=&title=Rethinking Weak-to-Strong Augmentation in Source-Free Domain Adaptive Object Detection&atitle=Rethinking%20Weak-to-Strong%20Augmentation%20in%20Source-Free%20Domain%20Adaptive%20Object%20Detection&aulast=Yang%2C%20Jiuzheng&id=DOI: Name: Full Text Finder (for New FTF UI) (s8985755) Category: fullText Text: Find It @ SCU Libraries MouseOverText: Find It @ SCU Libraries |
|---|---|
| Header | DbId: edsarx DbLabel: arXiv An: edsarx.2410.05557 RelevancyScore: 1112 AccessLevel: 3 PubType: Report PubTypeId: report PreciseRelevancyScore: 1112.26940917969 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Rethinking Weak-to-Strong Augmentation in Source-Free Domain Adaptive Object Detection – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Yang%2C+Jiuzheng%22">Yang, Jiuzheng</searchLink><br /><searchLink fieldCode="AR" term="%22Tang%2C+Song%22">Tang, Song</searchLink><br /><searchLink fieldCode="AR" term="%22Zhang%2C+Yangkuiyi%22">Zhang, Yangkuiyi</searchLink><br /><searchLink fieldCode="AR" term="%22Li%2C+Shuaifeng%22">Li, Shuaifeng</searchLink><br /><searchLink fieldCode="AR" term="%22Ye%2C+Mao%22">Ye, Mao</searchLink><br /><searchLink fieldCode="AR" term="%22Zhang%2C+Jianwei%22">Zhang, Jianwei</searchLink><br /><searchLink fieldCode="AR" term="%22Zhu%2C+Xiatian%22">Zhu, Xiatian</searchLink> – Name: DatePubCY Label: Publication Year Group: Date Data: 2024 – Name: Subset Label: Collection Group: HoldingsInfo Data: Computer Science – Name: Subject Label: Subject Terms Group: Su Data: <searchLink fieldCode="DE" term="%22Computer+Science+-+Computer+Vision+and+Pattern+Recognition%22">Computer Science - Computer Vision and Pattern Recognition</searchLink> – Name: Abstract Label: Description Group: Ab Data: Source-Free domain adaptive Object Detection (SFOD) aims to transfer a detector (pre-trained on source domain) to new unlabelled target domains. Current SFOD methods typically follow the Mean Teacher framework, where weak-to-strong augmentation provides diverse and sharp contrast for self-supervised learning. However, this augmentation strategy suffers from an inherent problem called crucial semantics loss: Due to random, strong disturbance, strong augmentation is prone to losing typical visual components, hindering cross-domain feature extraction. To address this thus-far ignored limitation, this paper introduces a novel Weak-to-Strong Contrastive Learning (WSCoL) approach. The core idea is to distill semantics lossless knowledge in the weak features (from the weak/teacher branch) to guide the representation learning upon the strong features (from the strong/student branch). To achieve this, we project the original features into a shared space using a mapping network, thereby reducing the bias between the weak and strong features. Meanwhile, a weak features-guided contrastive learning is performed in a weak-to-strong manner alternatively. Specifically, we first conduct an adaptation-aware prototype-guided clustering on the weak features to generate pseudo labels for corresponding strong features matched through proposals. Sequentially, we identify positive-negative samples based on the pseudo labels and perform cross-category contrastive learning on the strong features where an uncertainty estimator encourages adaptive background contrast. Extensive experiments demonstrate that WSCoL yields new state-of-the-art performance, offering a built-in mechanism mitigating crucial semantics loss for traditional Mean Teacher framework. The code and data will be released soon. – Name: TypeDocument Label: Document Type Group: TypDoc Data: Working Paper – Name: URL Label: Access URL Group: URL Data: <link linkTarget="URL" linkTerm="http://arxiv.org/abs/2410.05557" linkWindow="_blank">http://arxiv.org/abs/2410.05557</link> – Name: AN Label: Accession Number Group: ID Data: edsarx.2410.05557 |
| PLink | https://login.libproxy.scu.edu/login?url=https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&scope=site&db=edsarx&AN=edsarx.2410.05557 |
| RecordInfo | BibRecord: BibEntity: Subjects: – SubjectFull: Computer Science - Computer Vision and Pattern Recognition Type: general Titles: – TitleFull: Rethinking Weak-to-Strong Augmentation in Source-Free Domain Adaptive Object Detection Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Yang, Jiuzheng – PersonEntity: Name: NameFull: Tang, Song – PersonEntity: Name: NameFull: Zhang, Yangkuiyi – PersonEntity: Name: NameFull: Li, Shuaifeng – PersonEntity: Name: NameFull: Ye, Mao – PersonEntity: Name: NameFull: Zhang, Jianwei – PersonEntity: Name: NameFull: Zhu, Xiatian IsPartOfRelationships: – BibEntity: Dates: – D: 07 M: 10 Type: published Y: 2024 |
| ResultId | 1 |