Skip to content

Dataset · dronefreak

Composite Benchmark for Image Rain Removal (Rain13K)

dronefreak/Rain13K

Combined image dataset for training single-image deraining models.

What's inside

An unofficial mirror of the popular Rain13K training benchmark, including 10,000 to 100,000 images for image-to-image tasks. Five well-known datasets are combined inside: Rain800, Rain100H, Rain100L, Rain14000, and Rain12. The data is packaged for direct use in the ClearView pipeline.

What you can do with it

  • Training single-image rain removal (deraining) models
  • Reproducing SOTA architecture results like MPRNet and Restormer
  • Pre-training image enhancement and weather degradation removal models
Tasks
Image → image
Languages
English
Size
10K–100K rows
Downloads this month
44.3K
Likes
5

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("dronefreak/Rain13K")