Research map / 3D vision / Face and shape
Physically Realizable Natural-Looking Clothing Textures Evade Person Detectors via 3D Modeling
Zhanhao Hu, Wenda Chu, Xiaopei Zhu, Hui Zhang, Bo Zhang, Xiaolin Hu
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 AdvCaT instead of AdvPatch | Faster RCNN digital attack · Faster RCNN digital, IoU 0.1 | ASR (%): 72.27 -> 99.21 | Improved |
| use AdvCaT instead of AdvTexture | Faster RCNN digital attack · Faster RCNN digital, IoU 0.1 | ASR (%): 21.73 -> 99.21 | Improved |
| evaluate with the stricter IoU threshold 0.01 rather than 0.5 | AdvCaT · Faster RCNN digital | ASR (%): 95.18 -> 99.52 | No clear effect |
| use AdvCaT against the detector it targets | Faster RCNN, YOLOv3, and Deformable DETR · digital world | ASR (%): 99.36; 94.53; 88.77 | Mixed |
| transfer a Faster RCNN-targeted AdvCaT to other detectors | AdvCaT trained against Faster RCNN · Mask RCNN and YOLOv3 transfer | ASR (%): 92.22%; 23.26% | Mixed |
| add TopoProj and 3D TPS mesh augmentation during training (ablation) | AdvCaT · Faster RCNN physical | ASR (%): 19.27% -> 85.94% | Improved |
| add TopoProj augmentation during optimization (ablation) | AdvCaT · Faster RCNN under increasing mesh deformation | ASR | Improved |
| add 3D TPS augmentation during optimization (ablation) | AdvCaT · Faster RCNN under increasing mesh deformation | ASR | Improved |
| combine 3D TPS with TopoProj rather than using TopoProj alone (ablation) | AdvCaT · Faster RCNN under mesh deformation and unseen 3D models | ASR | Improved |
| increase the trainable-variable ratio lambda (ablation) | AdvCaT Gumbel seed parameterization · Faster RCNN digital | ASR | Improved |
| increase lambda (ablation) | AdvCaT camouflage appearance · camouflage-pattern naturalness | naturalness | Hurt |
| vary the color combination (ablation) | AdvCaT · Faster RCNN digital | ASR (%): >97% | No clear effect |
Closest papers by meaning
- Adversarial Texture for Fooling Person Detectors in the Physical World (CVPR 2022 · similarity 0.92)
- Physical Adversarial Examples for Person Detectors in Thermal Images Based on 3D Modeling (TPAMI 2025 · similarity 0.89)
- Design2Cloth: 3D Cloth Generation from 2D Masks (CVPR 2024 · similarity 0.87)
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.86)
- CNCA: Toward Customizable and Natural Generation of Adversarial Camouflage for Vehicle Detectors (NeurIPS 2024 · similarity 0.86)
- SCULPT: Shape-Conditioned Unpaired Learning of Pose-dependent Clothed and Textured Human Meshes (CVPR 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.
