Erosion Attack for Adversarial Training to Enhance Semantic Segmentation Robustness
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914269601202176 |
|---|---|
| author | Song, Yufei Zhou, Ziqi Deng, Menghao Hu, Yifan Hu, Shengshan Li, Minghui Zhang, Leo Yu |
| author_facet | Song, Yufei Zhou, Ziqi Deng, Menghao Hu, Yifan Hu, Shengshan Li, Minghui Zhang, Leo Yu |
| contents | Existing segmentation models exhibit significant vulnerability to adversarial attacks.To improve robustness, adversarial training incorporates adversarial examples into model training. However, existing attack methods consider only global semantic information and ignore contextual semantic relationships within the samples, limiting the effectiveness of adversarial training. To address this issue, we propose EroSeg-AT, a vulnerability-aware adversarial training framework that leverages EroSeg to generate adversarial examples. EroSeg first selects sensitive pixels based on pixel-level confidence and then progressively propagates perturbations to higher-confidence pixels, effectively disrupting the semantic consistency of the samples. Experimental results show that, compared to existing methods, our approach significantly improves attack effectiveness and enhances model robustness under adversarial training. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_14950 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Erosion Attack for Adversarial Training to Enhance Semantic Segmentation Robustness Song, Yufei Zhou, Ziqi Deng, Menghao Hu, Yifan Hu, Shengshan Li, Minghui Zhang, Leo Yu Computer Vision and Pattern Recognition Existing segmentation models exhibit significant vulnerability to adversarial attacks.To improve robustness, adversarial training incorporates adversarial examples into model training. However, existing attack methods consider only global semantic information and ignore contextual semantic relationships within the samples, limiting the effectiveness of adversarial training. To address this issue, we propose EroSeg-AT, a vulnerability-aware adversarial training framework that leverages EroSeg to generate adversarial examples. EroSeg first selects sensitive pixels based on pixel-level confidence and then progressively propagates perturbations to higher-confidence pixels, effectively disrupting the semantic consistency of the samples. Experimental results show that, compared to existing methods, our approach significantly improves attack effectiveness and enhances model robustness under adversarial training. |
| title | Erosion Attack for Adversarial Training to Enhance Semantic Segmentation Robustness |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.14950 |