You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Real Image Restoration Dataset

This dataset contains 9734 reproducibly sampled anime images for image restoration.

Each row contains:

  • ref1: a degraded full-image reference generated from the original image, saved as WebP by default.
  • target: the original clean image, saved as lossless WebP by default.
  • edit_instruction: an English instruction asking the model to restore clarity.
  • metadata columns: source paths, captions, descriptions, degradation settings, dimensions, and seeds.

The dataset was generated with seed 251100.

Hugging Face layout

train/
  ref1/
  prompt/
  target/
  metadata.jsonl
README.md

metadata.jsonl uses ref1_file_name and target_file_name so Hugging Face ImageFolder can display two image columns in the Dataset Viewer.

This dataset is generated by CPU image processing only. It does not use model inference. WebP is used by default to reduce storage while keeping Hugging Face Dataset Viewer compatibility.

Downloads last month
-