mayocream / anime-text-yolo

Anime Text YOLO

YOLO model weights for detecting anime text blocks. The repository is flattened and contains safetensors model files only, plus this model card and license.

Model Details

• Task: object detection
• Class label: text_block
• Format: safetensors PyTorch state dict
• Source format: converted from Ultralytics YOLO .pt checkpoints
• Metadata: each safetensors file stores class names, YOLO architecture YAML, training args, source path, conversion format, date, and Ultralytics version in its header

Files

Usage

These files are raw model state dictionaries, not pickled Ultralytics checkpoint objects. Instantiate the matching YOLOv12 model definition, then load the state dict.

from safetensors.torch import load_file

state_dict = load_file("yolo12n_animetext.safetensors")

Intended Use

Use these weights for detecting rectangular anime text block regions in images, such as manga, anime screenshots, or illustration datasets that contain rendered text.

Limitations

The repository does not include the original training data, evaluation artifacts, or ONNX exports. Users should validate performance on their own target image distribution before using the models in production.

License

This repository is licensed under GPL-3.0. See LICENSE .