Semantic Robustness Probing via Inpainting An Interactive Tool for Data Augmentation for Safety-Critical Object Detection
Nico Steckhan, Krutarth Prajapati, Weijia Shao, Jonas Heinle, Silvia Vock · 2026
Read Semantic Robustness Probing via Inpainting An Interactive Tool for Data Augmentation for Safety-Critical Object Detection by Nico Steckhan, Krutarth Prajapati, Weijia Shao, Jonas Heinle, Silvia Vock (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-2605-27155.txt.
Description
Steckhan and three co-authors present SemProbe, a tool for probing object finders in safety domains, beyond pixel damage. Users upload images, create masks by hand or by machine, pick design factors or custom prompts, and run diffusion-based controlled refill. The tool supports batch jobs, varied runs, and tunable settings. After each output, model scoring runs and shows labeled before-and-after views with score deltas, with each probe logged for traceable evidence. The team shows SemProbe on hand spotting for table saws, aimed at insurance test rules.
Catalog record
| CREATOR | Nico Steckhan, Krutarth Prajapati, Weijia Shao, Jonas Heinle, Silvia Vock |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 7891 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.27155