Research map / Adversarial robustness and security / Attacks on text and graphs
Attacking Optical Flow
Anurag Ranjan, Joel Janai, Andreas Geiger, Michael J. Black
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 |
|---|---|---|---|
| attack with a small adversarial patch (25x25, 0.1% of image) | FlowNetC · KITTI 2015 (white-box) | EPE: 14.56 -> 29.07 | Hurt |
| increase adversarial patch size to 153x153 (4.8%) | FlowNetC · KITTI 2015 (white-box) | EPE: 14.56 -> 95.32 | Hurt |
| attack with a 153x153 adversarial patch | FlowNet2 · KITTI 2015 (white-box) | EPE: 11.90 -> 59.58 | Hurt |
| attack with a 153x153 adversarial patch | SpyNet · KITTI 2015 (white-box) | EPE: 20.26 -> 21.00 | No clear effect |
| attack with a 153x153 adversarial patch | PWC-Net · KITTI 2015 (white-box) | EPE: 11.03 -> 12.52 | Mixed |
| attack with a 153x153 adversarial patch | Back2Future · KITTI 2015 (white-box) | EPE: 17.49 -> 18.43 | No clear effect |
| increase patch size from 25x25 to 153x153 for FlowNet2 (ablation) | FlowNet2 · KITTI 2015 (white-box) | EPE: 17.04 -> 59.58 | Hurt |
| attack with universal patch trained on FlowNet2 and PWC-Net | FlowNetC · KITTI 2015 (black-box) | EPE: 14.56 -> 86.12 | Hurt |
| attack with universal patch (unseen network) | FlowNet2 · KITTI 2015 (black-box) | EPE: 11.90 -> 36.13 | Hurt |
| attack with universal patch trained on FlowNet2 and PWC-Net | PWC-Net · KITTI 2015 (black-box) | EPE: 11.03 -> 11.01 | No clear effect |
| attack with universal patch (unseen classical method) | EpicFlow · KITTI 2015 (black-box) | EPE: 4.52 -> 4.66 | No clear effect |
| attack with universal patch (unseen classical method) | LDOF · KITTI 2015 (black-box) | EPE: 9.20 -> 9.17 | No clear effect |
Closest papers by meaning
- Towards Understanding Adversarial Robustness of Optical Flow Networks (CVPR 2022 · similarity 0.92)
- A Perturbation-Constrained Adversarial Attack for Evaluating the Robustness of Optical Flow (ECCV 2022 · similarity 0.90)
- Adversarial Imaging Pipelines (CVPR 2021 · similarity 0.88)
- Physical Adversarial Attack Meets Computer Vision: A Decade Survey (TPAMI 2024 · similarity 0.87)
- With Friends Like These, Who Needs Adversaries? (NeurIPS 2018 · similarity 0.86)
- BadPart: Unified Black-box Adversarial Patch Attacks against Pixel-wise Regression Tasks (ICML 2024 · similarity 0.85)
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.
