Research map / Adversarial robustness and security / Attacks on text and graphs
Cooling-Shrinking Attack: Blinding the Tracker With Imperceptible Noises
Bin Yan, Dong Wang, Huchuan Lu, Xiaoyun Yang
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add cooling-shrinking perturbations to search regions | SiamRPNpp · OTB100 | Success: 0.696 -> 0.349 | Hurt |
| Add cooling-shrinking perturbations to search regions | SiamRPNpp · VOT2018 | EAO: 0.414 -> 0.073 | Hurt |
| Add cooling-shrinking perturbations to search regions | SiamRPNpp · LaSOT | Success: 0.496 -> 0.180 | Hurt |
| Perturb the template once in the initial frame | SiamRPNpp · OTB100 | Success: 0.696 -> 0.527 | Hurt |
| Attack both the template and search regions | SiamRPNpp · OTB100 | Success: 0.696 -> 0.324 | Hurt |
| Add shrinking loss (ablation) | Search-region attack · OTB100 and LaSOT | Attack success | Improved |
| Add shrinking loss (ablation) | Template-only attack · OTB100 and LaSOT | Attack success | Hurt |
| Use an L2 loss instead of adding a discriminator (ablation) | Cooling-shrinking perturbation generator · Tracking attack | Imperceptibility | No clear effect |
| Transfer the search-region attack to DaSiamRPN | DaSiamRPN · LaSOT | Success: 0.458 -> 0.400 | Hurt |
| Transfer the search-region attack to DiMP50 | DiMP50 · LaSOT | Success: 0.559 -> 0.492 | Hurt |
| Replace learned perturbations with impulse or Gaussian noise | SiamRPNpp search regions · OTB100 | Success: 0.349 vs 0.486, 0.389, 0.553, 0.389 | Mixed |
Closest papers by meaning
- Efficient Adversarial Attacks for Visual Object Tracking (ECCV 2020 · similarity 0.90)
- One-Shot Adversarial Attacks on Visual Tracking With Dual Attention (CVPR 2020 · similarity 0.88)
- IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking (CVPR 2021 · similarity 0.87)
- Robust Tracking against Adversarial Attacks (ECCV 2020 · similarity 0.87)
- Enhancing Tracking Robustness with Auxiliary Adversarial Defense Networks (ECCV 2024 · similarity 0.85)
- SlowFormer: Adversarial Attack on Compute and Energy Consumption of Efficient Vision Transformers (CVPR 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
