PatchPoison: Poisoning Multi-View Datasets to Degrade 3D Reconstruction
Prajas Wadekar, Venkata Sai Pranav Bachina, Kunal Bhosikar, Charu Sharma, [email protected] · 2026
Read PatchPoison: Poisoning Multi-View Datasets to Degrade 3D Reconstruction by Prajas Wadekar, Venkata Sai Pranav Bachina, Kunal Bhosikar, Charu Sharma, [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-13153.txt.
Description
Wadekar and four colleagues present PatchPoison, a lightweight poisoning method that stops unauthorized 3D reconstruction from public photos and videos. It inserts a small high-frequency checkerboard patch at the edge of each image, corrupting feature matching in Structure-from-Motion pipelines so estimated camera poses misalign and downstream 3D optimization drifts off true geometry. On the NeRF-Synthetic benchmark, a 12 by 12 pixel patch raises reconstruction error 6.8 times in LPIPS while staying unobtrusive to viewers.
Catalog record
| CREATOR | Prajas Wadekar, Venkata Sai Pranav Bachina, Kunal Bhosikar, Charu Sharma, [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36821 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.13153