Anime Image Segmentation Dataset with 26,000 Foreground and Background Images

26,000 anime-style images curated for segmentation tasks, split evenly between foreground subjects and backgrounds. The dataset was created by Zarxrax through automated and manual inspection to improve upon a foundational source. It was last updated on January 28, 2024.

Use Cases

• Train foreground-background segmentation models based on the paired image structure.
• Develop image matting algorithms for anime characters based on the described foreground/background split.
• Benchmark performance of segmentation models on stylized, non-photographic imagery.
• Create synthetic training data for anime-style image generation pipelines.

Strengths

• Contains 26,000 total images, providing a substantial volume for model training.
• Explicitly contains 13,000 foreground images and 13,000 background images, enabling supervised segmentation tasks.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2024-01-28 16:42:00; freshness should be verified.

Provenance

Source

Derived from the skytnt/anime-segmentation dataset on Hugging Face.

Collection Method

Created through automated and manual inspection and curation by the author Zarxrax.

Freshness

Last updated 2024-01-28 16:42:00