README.md

metadata

pretty_name: Anime Orientation Correction Dataset
task_categories:
  - image-to-image
language:
  - en
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files: train/metadata.jsonl

Anime Orientation Correction Dataset

This dataset contains 9734 reproducibly sampled anime images for image orientation correction.

Each row contains:

• ref1: the original image rotated by 90, 180, or 270 degrees.
• target: the original image in its correct upright orientation.
• edit_instruction: an English instruction asking the model to correct the image orientation.
• metadata columns containing source paths, captions, descriptions, rotation angle, dimensions, and seed.

The dataset was generated with seed 251100 using CPU image processing only.

Hugging Face layout

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

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