Research map / Detection and segmentation / Tracking
Learning To Adversarially Blur Visual Object Tracking
Qing Guo, Ziyi Cheng, Felix Juefei-Xu, Lei Ma, Xiaofei Xie, Yang Liu, Jianjun Zhao
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 |
|---|---|---|---|
| tune accumulation weights A_t in addition to motion W_t (ablation) | OS-ABA · OTB100 (SiamRPN++ ResNet50) | success rate drop: 5.5 -> 28.1 | Improved |
| tune accumulation weights A_t in addition to motion W_t (ablation) | OP-ABA · OTB100 (SiamRPN++ ResNet50) | success rate drop: 15.0 -> 31.2 | Improved |
| tune motion ratios W_t in addition to accumulation A_t (ablation) | OP-ABA · OTB100 (SiamRPN++ ResNet50) | success rate drop: 25.6 -> 31.2 | Improved |
| tune motion ratios W_t in addition to accumulation A_t (ablation) | OS-ABA · OTB100 (SiamRPN++ ResNet50) | success rate drop: 24.9 -> 28.1 | Improved |
| tune motion ratios W_t in addition to accumulation A_t (ablation) | OS-ABA · OTB100 (SiamRPN++ ResNet50) | precision drop: 29.5 -> 32.5 | Improved |
| use adversarial blur instead of normal (Norm-Blur) blur | SiamRPN++ ResNet50 · OTB100 | success rate drop: 1.2 -> 31.2 | Improved |
| use iterative optimization (OP-ABA) instead of one-step network (OS-ABA) | adversarial blur attack · OTB100 (SiamRPN++ ResNet50) | precision drop: 32.5 -> 41.7 | Improved |
| use one-step JAMANet (OS-ABA) instead of optimization (OP-ABA) for speed | adversarial blur attack · OTB100 (SiamRPN++ ResNet50) | attack FPS: 6.79 -> 17.62 | Improved |
| use lighter backbone MobileNetv2 instead of ResNet50 as target | OP-ABA · OTB100 | precision drop: 41.7 -> 49.6 | Mixed |
| use OS-ABA instead of OP-ABA for transfer to KYS | SiamRPN++ ResNet50 guided attack · OTB100 (transfer) | precision drop: 7.9 -> 12.3 | Improved |
| use OS-ABA instead of CSA for transfer to DiMP50 | SiamRPN++ ResNet50 guided attack · OTB100 (transfer) | precision drop: 3.4 -> 10.7 | Improved |
| use OS-ABA instead of OP-ABA for transfer to SiamRPN++ MobileNetv2 | SiamRPN++ ResNet50 guided attack · OTB100 (transfer) | precision drop: 2.5 -> 0.2 | Hurt |
Closest papers by meaning
- Watch out! Motion is Blurring the Vision of Your Deep Neural Networks (NeurIPS 2020 · similarity 0.90)
- IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking (CVPR 2021 · similarity 0.88)
- Robust Tracking against Adversarial Attacks (ECCV 2020 · similarity 0.88)
- Physical Adversarial Textures That Fool Visual Object Tracking (ICCV 2019 · similarity 0.88)
- DeTrack: In-model Latent Denoising Learning for Visual Object Tracking (NeurIPS 2024 · similarity 0.85)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 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.
