---
license: openrail
task_categories:
- image-classification
tags:
- art
size_categories:
- 100K<n<1M
---

The dataset is used for classifying portrait images in anime videos and is specifically divided into the following four categories:

* **Vision**: Overall character images (may include instances of facial drawing issues; not recommended to split when training LoRA).
* **Imagery**: Clear character images (both facial and upper-body features are clear).
* **Halfbody**: Upper-body character images (mostly dominated by the head, with some partial upper-body features).
* **Face**: Close-up shots of character faces (close-up shots of faces with highly clear features in appearance and expression).

| Dataset | Vision | Imagery | Halfbody | Face  |
|---------|--------|---------|----------|-------|
| v0      | 46866  | 49671   | 49685    | 49438 |