Model weights for executing the mangaka cli.
The models are ONNX exports of various checkpoints.
usage
- the sam models are for bubble speech detection
- the ocr models are for character recognition
- dbconvnext is used for text box recognition
- lama is for impatingin
licensing
The SAM models are under the SAM license, and the other models are covered under their appropriate exports. https://github.com/facebookresearch/sam3/blob/main/LICENSE
The models otherwise are otherwise not licensable properly since its all documented as model weights, which aren't necessarily code, so assume CC-BY?
- dbconvnext (https://github.com/zyddnys/manga-image-translator/) - GPL
- ocr-48px (https://github.com/zyddnys/manga-image-translator/) - GPL
- lama (https://huggingface.co/Carve/LaMa-ONNX) - Apache