Research map / Adversarial robustness and security / Black-box and transfer attacks
Merry Go Round: Rotate a Frame and Fool a DNN
Daksh Thapar, Aditya Nigam, Chetan Arora
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 |
|---|---|---|---|
| replace ME-Sampler with the proposed mixed intensity-and-rotation attack | I3D · Kinetics-400 untargeted | ANQ: 2,717 -> 1,257 | Improved |
| replace ME-Sampler with the proposed mixed intensity-and-rotation attack | Rolling-Unrolling LSTM · Epic-Kitchens untargeted | ANQ: 7,326 -> 3,564 | Improved |
| replace V-Bad with the proposed mixed intensity-and-rotation attack | EgoGaitNet · IITMD-FPR untargeted | ANQ: 5,480 -> 3,487 | Improved |
| replace ME-Sampler with the proposed mixed intensity-and-rotation attack | I3D · Kinetics-400 targeted | ANQ: 11,120 -> 6,234 | Improved |
| replace ME-Sampler with the proposed mixed intensity-and-rotation attack | Rolling-Unrolling LSTM · Epic-Kitchens targeted | ANQ: 22,541 -> 15,283 | Improved |
| replace ME-Sampler with the proposed mixed intensity-and-rotation attack | EgoGaitNet · IITMD-FPR targeted | ANQ: 18,759 -> 9,910 | Improved |
| use camera rotation only instead of the learned mixed perturbation (ablation) | I3D attack · Kinetics-400 | SR%: 99.33 -> 38.19 | Hurt |
| use camera rotation only instead of the learned mixed perturbation (ablation) | Rolling-Unrolling LSTM attack · Epic-Kitchens | SR%: 100.00 -> 62.81 | Hurt |
| use camera rotation only instead of the learned mixed perturbation (ablation) | EgoGaitNet attack · IITMD-FPR | SR%: 96.33 -> 58.42 | Hurt |
| use intensity-based perturbation only instead of the learned mixture (ablation) | I3D attack · Kinetics-400 | ANQ: 1,257 -> 3,569 | Hurt |
| use intensity-based perturbation only instead of the learned mixture (ablation) | Rolling-Unrolling LSTM attack · Epic-Kitchens | ANQ: 3,564 -> 8,238 | Hurt |
| use intensity-based perturbation only instead of the learned mixture (ablation) | EgoGaitNet attack · IITMD-FPR | ANQ: 3,487 -> 6,356 | Hurt |
Closest papers by meaning
- Sparse Adversarial Perturbations for Videos (AAAI 2019 · similarity 0.87)
- AdvIT: Adversarial Frames Identifier Based on Temporal Consistency in Videos (ICCV 2019 · similarity 0.86)
- Catastrophic Child's Play: Easy to Perform, Hard to Defend Adversarial Attacks (CVPR 2019 · similarity 0.86)
- Physical Adversarial Attack Meets Computer Vision: A Decade Survey (TPAMI 2024 · similarity 0.86)
- Breaking Temporal Consistency: Generating Video Universal Adversarial Perturbations Using Image Models (ICCV 2023 · similarity 0.85)
- Object-Shot Enhanced Grounding Network for Egocentric Video (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.
