Research map / Adversarial robustness and security / Attacks on text and graphs
Initialization and Alignment for Adversarial Texture Optimization
Xiaoming Zhao, Zhizhen Zhao, Alexander G. Schwing
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 |
|---|---|---|---|
| add TexSmooth (adversarial optimization with FFT alignment) after TexInit (ablation) | TexInit · UofI Texture Scenes (11 scenes) | LPIPS: 0.342 -> 0.309 | Improved |
| add TexSmooth after TexInit (ablation) | TexInit · UofI Texture Scenes (11 scenes) | SSIM: 0.510 -> 0.602 | Improved |
| add TexInit instead of L2Avg initialization (ablation) | Ours · UofI Texture Scenes (11 scenes) | LPIPS: 0.332 -> 0.309 | Improved |
| add TexInit instead of L2Avg initialization (ablation) | Ours · UofI Texture Scenes (11 scenes) | SSIM: 0.592 -> 0.602 | Improved |
| add FFT alignment to AdvOptim (ablation) | Ours · UofI Texture Scenes (11 scenes) | LPIPS: 0.346 -> 0.309 | Improved |
| add FFT alignment to AdvOptim (ablation) | Ours · UofI Texture Scenes (11 scenes) | SSIM: 0.559 -> 0.602 | Improved |
| add pairwise adjacency cues to TexInit (ablation) | Ours · UofI Texture Scenes (11 scenes) | LPIPS: 0.309 -> 0.305 | No clear effect |
| split alignment into 2x2 patches (ablation) | Ours · UofI Texture Scenes (11 scenes) | LPIPS: 0.309 -> 0.330 | Hurt |
| split alignment into 4x4 patches (ablation) | Ours · UofI Texture Scenes (11 scenes) | SSIM: 0.602 -> 0.545 | Hurt |
| train with sparser views (best vs worst of k=1..5) (ablation) | Ours · UofI Texture Scenes (11 scenes) | SSIM: 0.604 -> 0.580 | Hurt |
| use Ours instead of AdvTex-C under noisy camera poses | AdvTex-C · UofI Texture Scenes (11 scenes) (noisy camera poses) | LPIPS: 0.490 -> 0.472 | Improved |
| use Ours instead of AdvTex-B | AdvTex-B · ScanNet (scenes ID <= 20) | LPIPS: 0.557 -> 0.503 | Improved |
Closest papers by meaning
- Adversarial Texture Optimization From RGB-D Scans (CVPR 2020 · similarity 0.89)
- Fine Detailed Texture Learning for 3D Meshes With Generative Models (TPAMI 2023 · similarity 0.88)
- AUV-Net: Learning Aligned UV Maps for Texture Transfer and Synthesis (CVPR 2022 · similarity 0.87)
- Learning Texture Manifolds with the Periodic Spatial GAN (ICML 2017 · similarity 0.86)
- Learning Generative Models of Textured 3D Meshes From Real-World Images (ICCV 2021 · similarity 0.86)
- GramGAN: Deep 3D Texture Synthesis From 2D Exemplars (NeurIPS 2020 · 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.
