Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval

Qiwei Tian, Chenhao Lin, Zhengyu Zhao, Qian Li, Chao Shen
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:48139-48153, 2024.

Abstract

Adversarial training has achieved substantial performance in defending image retrieval against adversarial examples. However, existing studies in deep metric learning (DML) still suffer from two major limitations: weak adversary and model collapse. In this paper, we address these two limitations by proposing Collapse-Aware TRIplet DEcoupling (CA-TRIDE). Specifically, TRIDE yields a stronger adversary by spatially decoupling the perturbation targets into the anchor and the other candidates. Furthermore, CA prevents the consequential model collapse, based on a novel metric, collapseness, which is incorporated into the optimization of perturbation. We also identify two drawbacks of the existing robustness metric in image retrieval and propose a new metric for a more reasonable robustness evaluation. Extensive experiments on three datasets demonstrate that CA-TRIDE outperforms existing defense methods in both conventional and new metrics. Codes are available at https://github.com/michaeltian108/CA-TRIDE.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-tian24a, title = {Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval}, author = {Tian, Qiwei and Lin, Chenhao and Zhao, Zhengyu and Li, Qian and Shen, Chao}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {48139--48153}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/tian24a/tian24a.pdf}, url = {https://proceedings.mlr.press/v235/tian24a.html}, abstract = {Adversarial training has achieved substantial performance in defending image retrieval against adversarial examples. However, existing studies in deep metric learning (DML) still suffer from two major limitations: weak adversary and model collapse. In this paper, we address these two limitations by proposing Collapse-Aware TRIplet DEcoupling (CA-TRIDE). Specifically, TRIDE yields a stronger adversary by spatially decoupling the perturbation targets into the anchor and the other candidates. Furthermore, CA prevents the consequential model collapse, based on a novel metric, collapseness, which is incorporated into the optimization of perturbation. We also identify two drawbacks of the existing robustness metric in image retrieval and propose a new metric for a more reasonable robustness evaluation. Extensive experiments on three datasets demonstrate that CA-TRIDE outperforms existing defense methods in both conventional and new metrics. Codes are available at https://github.com/michaeltian108/CA-TRIDE.} }
Endnote
%0 Conference Paper %T Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval %A Qiwei Tian %A Chenhao Lin %A Zhengyu Zhao %A Qian Li %A Chao Shen %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-tian24a %I PMLR %P 48139--48153 %U https://proceedings.mlr.press/v235/tian24a.html %V 235 %X Adversarial training has achieved substantial performance in defending image retrieval against adversarial examples. However, existing studies in deep metric learning (DML) still suffer from two major limitations: weak adversary and model collapse. In this paper, we address these two limitations by proposing Collapse-Aware TRIplet DEcoupling (CA-TRIDE). Specifically, TRIDE yields a stronger adversary by spatially decoupling the perturbation targets into the anchor and the other candidates. Furthermore, CA prevents the consequential model collapse, based on a novel metric, collapseness, which is incorporated into the optimization of perturbation. We also identify two drawbacks of the existing robustness metric in image retrieval and propose a new metric for a more reasonable robustness evaluation. Extensive experiments on three datasets demonstrate that CA-TRIDE outperforms existing defense methods in both conventional and new metrics. Codes are available at https://github.com/michaeltian108/CA-TRIDE.
APA
Tian, Q., Lin, C., Zhao, Z., Li, Q. & Shen, C.. (2024). Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:48139-48153 Available from https://proceedings.mlr.press/v235/tian24a.html.

Related Material