Research map / 3D vision / Depth and stereo
PoInit-of-View: Poisoning Initialization of Views Transfers Across Multiple 3D Reconstruction Systems
Weijie Wang, Songlong Xing, Zhengyu Zhao, Nicu Sebe, Bruno Lepri
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 cross-view gradient inconsistency instead of single-view loss | PoInit-of-View · Tanks and Temples, Mip-Splatting | PSNR: 23.53 -> 17.63 (lower PSNR means stronger attack) | Improved |
| use cross-view poisoning instead of adding Gaussian noise | PoInit-of-View · Tanks and Temples, Mip-Splatting | PSNR: 23.43 -> 17.63 | Improved |
| add SSIM regularizer to the proxy objective (ablation) | PoInit-of-View · Tanks and Temples, Mip-Splatting | downstream PSNR: 19.4 -> 17.63 | Improved |
| add TV regularizer to the proxy objective (ablation) | PoInit-of-View · Tanks and Temples, Mip-Splatting | downstream PSNR: 19.7 -> 17.63 | Improved |
| add SSIM regularizer to the proxy objective (ablation) | PoInit-of-View · Tanks and Temples, Mip-Splatting | clean-vs-poisoned PSNR (stealth): 24.8 -> 22.3 | Hurt |
| add cross-view inconsistency loss in place of photometric-only loss (ablation) | PoInit-of-View · Tanks and Temples, Mip-Splatting | registered views (%): 91.2 -> 24.3 | Improved |
| add cross-view inconsistency loss in place of photometric-only loss (ablation) | PoInit-of-View · Tanks and Temples, Mip-Splatting | downstream PSNR: 22.8 -> 17.63 | Improved |
| add SSIM regularizer to the proxy objective (ablation) | PoInit-of-View · Tanks and Temples, Mip-Splatting | registered views (%): 27.1 -> 24.3 | Improved |
| increase perturbation budget beyond 12/255 up to 32/255 (ablation) | PoInit-of-View · Tanks and Temples | registration rate, PSNR, SSIM: sharp drop once rho exceeds 12/255; almost complete failure at 32/255 | Improved |
| poison inputs with PoInit-of-View versus clean inputs | Colmap · Tanks and Temples | PSNR: 11.92 -> 8.96 | Improved |
| poison inputs with PoInit-of-View versus clean inputs | Instant NGP · Tanks and Temples | PSNR: 21.62 -> 16.24 | Improved |
| poison inputs with PoInit-of-View versus clean inputs | Mip-Splatting · Tanks and Temples | LPIPS: 0.166 -> 0.327 | Improved |
Closest papers by meaning
- Understanding Multi-View Transformers (NeurIPS 2026 · similarity 0.86)
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 2025 · similarity 0.85)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.84)
- Leveraging Camera Triplets for Efficient and Accurate Structure-from-Motion (CVPR 2024 · similarity 0.84)
- View Transformation Robustness for Multi-View 3D Object Reconstruction with Reconstruction Error-Guided View Selection (AAAI 2025 · similarity 0.84)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · 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.
