Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Physically Realizable Natural-Looking Clothing Textures Evade Person Detectors via 3D Modeling

Zhanhao Hu, Wenda Chu, Xiaopei Zhu, Hui Zhang, Bo Zhang, Xiaolin Hu

TL;DR

This work proposes adversarial camou-flage textures (AdvCaT) that resemble one kind of the typical textures of daily clothes, camouflage textures and proposes an efficient augmentation pipeline on 3D meshes combining topologically plausible projection (TopoProj) and Thin Plate Spline (TPS) to narrow the gap between digital and real-world objects.

Abstract

Recent works have proposed to craft adversarial clothes for evading person detectors, while they are either only effective at limited viewing angles or very conspicuous to humans. We aim to craft adversarial texture for clothes based on 3D modeling, an idea that has been used to craft rigid adversarial objects such as a 3D-printed turtle. Unlike rigid objects, humans and clothes are non-rigid, leading to difficulties in physical realization. In order to craft natural-looking adversarial clothes that can evade person detectors at multiple viewing angles, we propose adversarial camouflage textures (AdvCaT) that resemble one kind of the typical textures of daily clothes, camouflage textures. We leverage the Voronoi diagram and Gumbel-softmax trick to parameterize the camouflage textures and optimize the parameters via 3D modeling. Moreover, we propose an efficient augmentation pipeline on 3D meshes combining topologically plausible projection (TopoProj) and Thin Plate Spline (TPS) to narrow the gap between digital and real-world objects. We printed the developed 3D texture pieces on fabric materials and tailored them into T-shirts and trousers. Experiments show high attack success rates of these clothes against multiple detectors.

Physically Realizable Natural-Looking Clothing Textures Evade Person Detectors via 3D Modeling

TL;DR

This work proposes adversarial camou-flage textures (AdvCaT) that resemble one kind of the typical textures of daily clothes, camouflage textures and proposes an efficient augmentation pipeline on 3D meshes combining topologically plausible projection (TopoProj) and Thin Plate Spline (TPS) to narrow the gap between digital and real-world objects.

Abstract

Recent works have proposed to craft adversarial clothes for evading person detectors, while they are either only effective at limited viewing angles or very conspicuous to humans. We aim to craft adversarial texture for clothes based on 3D modeling, an idea that has been used to craft rigid adversarial objects such as a 3D-printed turtle. Unlike rigid objects, humans and clothes are non-rigid, leading to difficulties in physical realization. In order to craft natural-looking adversarial clothes that can evade person detectors at multiple viewing angles, we propose adversarial camouflage textures (AdvCaT) that resemble one kind of the typical textures of daily clothes, camouflage textures. We leverage the Voronoi diagram and Gumbel-softmax trick to parameterize the camouflage textures and optimize the parameters via 3D modeling. Moreover, we propose an efficient augmentation pipeline on 3D meshes combining topologically plausible projection (TopoProj) and Thin Plate Spline (TPS) to narrow the gap between digital and real-world objects. We printed the developed 3D texture pieces on fabric materials and tailored them into T-shirts and trousers. Experiments show high attack success rates of these clothes against multiple detectors.
Paper Structure (25 sections, 12 equations, 20 figures, 5 tables)

This paper contains 25 sections, 12 equations, 20 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Adversarial Camouflage Texture Patterns Generation
  4. Differentiable Generation of Camouflage Textures
  5. Non-rigid Mesh Augmentations
  6. Adversarial Loss Function
  7. Experiments
  8. Experimental Setup
  9. Naturalness Score by Subjective Evaluation
  10. Digital World Results
  11. Physical World Results
  12. Conclusion
  13. Supplementary Methods
  14. Topologically Plausible Projection Generation
  15. TPS Warping on TopoProj
  16. ...and 10 more sections

Figures (20)

  • Figure 1: Visualization of several adversarial clothes. (a) Adversarial patch thys2019fooling. (b) Adversarial T-shirt xu2020adversarial. (c) Naturalistic patch hu2021naturalistic. (d) Adversarial Texture hu2022adversarial. (e) Left: daily camouflage texture; Right: our adversarial camouflage texture.
  • Figure 2: The training pipeline of the adversarial camouflage textures.
  • Figure 3: (a) Camouflage texture. (b) Color cluster regions. The region of each color cluster can be approximately represented by polygons. (c) Voronoi diagram. The blue points are the control points, and the red lines are the boundaries of the regions.
  • Figure 4: Visualization of the texture augmentation. (a) The texture map of a T-shirt mesh. It is geometrically plausible. (b) A texture map that is topologically plausible, where each point's neighbors in the 2D projection correspond precisely to its neighbors in the 3D mesh. (c-e) Rendered images with different warp methods on the texture map. (c) No warp. (d) Applying a mild shear strain along the texture map's vertical axis. The red arrows indicate pixels that appear at wrong places on the rendered image. (e) Our texture warp base on TopoProj.
  • Figure 5: Illustration of the warping at the edge of two pieces. A GeoProj and a TopoProj with textures are shown on the left panel. The two pieces are far away on GeoProj but next to each other on TopoProj. Two triangle elements with blue solid lines at the eadges of the pieces are next to each other on TopoProj. The vertices on TopoProj and GeoProj with the same color are the projections of an identical vertex on the 3D mesh.
  • ...and 15 more figures