Research map / Adversarial robustness and security / Attacks on text and graphs
SPARK: Spatial-aware Online Incremental Attack Against Visual Tracking
Qing Guo, Xiaofei Xie, Felix Juefei-Xu, Lei Ma, Zhongguo Li, Wanli Xue, Wei Feng, Yang Liu
Read the paperCheck 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 |
|---|---|---|---|
| use SPARK incremental attack instead of BA-E with BIM | SiamRPN-AlexNet · OTB100 TA | success rate: 69.6 -> 78.9 | Improved |
| attack every 10th frame and reuse perturbation (BA-R2) instead of random 10% frames (BA-R1), BIM | SiamRPN-AlexNet · OTB100 TA | success rate: 17.8 -> 53.7 | Improved |
| reuse perturbations across frames (BA-R1 vs BA-E with BIM) | SiamRPN-AlexNet · OTB100 TA | success rate: 69.6 -> 17.8 | Hurt |
| add SPARK incremental perturbation with L2,1 norm instead of BIM per-frame | SiamRPN-AlexNet · OTB100 UA | precision drop: 72.1 -> 78.9 | Improved |
| add SPARK instead of BIM on ResNet50 | SiamRPN-ResNet50 · OTB100 UA | precision drop: 31.9 -> 29.8 | No clear effect |
| use SPARK instead of BIM on MobileNetv2 for targeted attack | SiamRPN-MobileNetv2 · OTB100 TA | success rate: 33.5 -> 51.4 | Improved |
| use given object template T instead of SSD-detected object (SPARK-noT) (ablation) | SPARK · OTB100 UA | precision drop: 71.0 -> 78.9 | Improved |
| use given object template T instead of SSD-detected object (SPARK-noT) (ablation) | SPARK · OTB100 TA | success rate: 50.6 -> 74.6 | Improved |
| use the attacked tracker's predictions instead of surrogate predictions (SPARK-no b) (ablation) | SPARK · OTB100 UA | precision drop: 67.7 -> 78.9 | Improved |
| use the attacked tracker's predictions instead of surrogate predictions (SPARK-no b) (ablation) | SPARK · OTB100 TA | success rate: 46.2 -> 74.6 | Improved |
| increase history length L from 5 to 30 in SPARK (ablation) | SPARK · OTB100 TA, SiamRPN-AlexNet | success rate | Improved |
| increase history length L (perturbation size) (ablation) | SPARK · OTB100 TA, SiamRPN-AlexNet | MAP | Mixed |
Closest papers by meaning
- IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking (CVPR 2021 · similarity 0.90)
- Efficient Adversarial Attacks for Visual Object Tracking (ECCV 2020 · similarity 0.89)
- One-Shot Adversarial Attacks on Visual Tracking With Dual Attention (CVPR 2020 · similarity 0.89)
- Towards Universal Physical Attacks on Single Object Tracking (AAAI 2021 · similarity 0.89)
- LRR: Language-Driven Resamplable Continuous Representation against Adversarial Tracking Attacks (ICLR 2024 · similarity 0.84)
- I2VGuard: Safeguarding Images against Misuse in Diffusion-based Image-to-Video Models (CVPR 2025 · similarity 0.84)
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.
