OpenRR-5k: A Large-Scale Benchmark for Reflection Removal in the Wild
Jie Cai, Kangning Yang, Ling Ouyang, Lan Fu, Jiaming Ding, Jinglin Shen, Zibo Meng · 2025
Read OpenRR-5k: A Large-Scale Benchmark for Reflection Removal in the Wild by Jie Cai, Kangning Yang, Ling Ouyang, Lan Fu, Jiaming Ding, Jinglin Shen, Zibo Meng (2025) 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-2506-05482.txt.
Description
Jie Cai and co-authors build OpenRR-5k, a large data set for teaching computers to remove reflections from photos. Their paper appeared in 2025. Reflections on glass spoil many photos, yet training data for fixing them has been small and narrow. The new set holds 5,300 matched pairs of photos, each pair showing the same scene with and without the reflection. Most pairs are for training, with smaller sets for checking and testing.
The photos cover many lighting setups, objects, and reflection styles, all lined up pixel by pixel. The team also added 100 real-world photos with no clean match, to test methods in the wild. They trained a standard model on the set and scored it with five common measures. The data set is meant to give the field a fair, shared test bed.
Catalog record
| CREATOR | Jie Cai, Kangning Yang, Ling Ouyang, Lan Fu, Jiaming Ding, Jinglin Shen, Zibo Meng |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 18 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-05482