Dataset · Mingde
Polarization-based reflection removal dataset (CVPR 2025)
Mingde/PolaRGB
Paired images for reflection removal and image enhancement based on polarization.
What's inside
The dataset contains between 1,000 and 10,000 images for image-to-image tasks presented in the PolarFree paper at CVPR 2025. The data is intended for research in polarization imaging and cleaning photographs of spurious reflections.
What you can do with it
- Training neural networks to remove glare and reflections from glass or water
- Evaluating the quality of image enhancement algorithms
- Tasks
- Image → image
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 70.6K
- Likes
- 2
How to load it
from datasets import load_dataset
ds = load_dataset("Mingde/PolaRGB")