Research map / 3D vision / Human pose estimation
Kaleidoscopic Background Attack: Disrupting Pose Estimation with Multi-Fold Radial Symmetry Textures
Xinlong Ding, Hongwei Yu, Jiawei Li, Feifan Li, Yu Shang, Bochao Zou, Huimin Ma, Jiansheng Chen
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 |
|---|---|---|---|
| build background as an N-fold radially symmetric kaleidoscopic disc from natural texture segments | DUSt3R · physical world | mAA(30): 0.95 -> 0.31 | Improved |
| optimize kaleidoscopic segments with projected orientation consistency loss | KBA_nat on DUSt3R · physical world | mAA(30): 0.31 -> 0.00 | Improved |
| transfer KBA_opt optimized on DUSt3R to black-box MASt3R | KBA_opt · physical world, MASt3R | mAA(30): 0.80 -> 0.05 | Improved |
| transfer KBA_opt to black-box RelPose++ | KBA_opt · physical world, RelPose++ | mAA(30): 0.43 -> 0.17 | Improved |
| use natural radially/translationally symmetric textures (woven, tiles) without kaleidoscopic construction (ablation) | DUSt3R · DT1 5 images | RRA@15: Nature 0.97 vs N-3 0.97 | No clear effect |
| construct kaleidoscopic disc from natural patterns (KN-3) instead of using the pattern directly (ablation) | natural texture N-3 · DT1 5 images, DUSt3R | RRA@15: 0.97 -> 0.45 | Improved |
| replace L_poc with output MSE loss across views (ablation) | KBA_opt · DT1 5 images, DUSt3R | mAA(30): 0.08 -> 0.13 | Hurt |
| use kaleidoscopic disc (KP) instead of whole-image patch (NP) with L_poc (ablation) | KBA_opt · DT1 5 images, DUSt3R | mAA(30): 0.29 -> 0.08 | Improved |
| use more views (3, 5, 10) at test time than the two used in optimization | KBA_opt · digital world | RRA/RTA/mAA | No clear effect |
Closest papers by meaning
- DepthVanish: Optimizing Adversarial Interval Structures for Stereo-Depth-Invisible Patches (NeurIPS 2025 · similarity 0.83)
- UV-Attack: Physical-World Adversarial Attacks on Person Detection via Dynamic-NeRF-based UV Mapping (ICLR 2025 · similarity 0.81)
- iG-6DoF: Model-free 6DoF Pose Estimation for Unseen Object via Iterative 3D Gaussian Splatting (CVPR 2025 · similarity 0.81)
- RUBIK: A Structured Benchmark for Image Matching across Geometric Challenges (CVPR 2025 · similarity 0.81)
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.81)
- Efficient Privacy-Preserving Visual Localization Using 3D Ray Clouds (CVPR 2024 · similarity 0.81)
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.
