iamokish/manga-text-segmentation-safetensors
Manga Text Segmentation 2025 — SafeTensors
This repository contains a SafeTensors conversion of the original PyTorch model from:
Original model: https://huggingface.co/a-b-c-x-y-z/Manga-Text-Segmentation-2025
The original model is designed for text segmentation in manga and uses a U-Net++ architecture with an EfficientNetV2 backbone.
Conversion
The original model weights were provided in PyTorch .pth format.
They were converted to `.safetensors` format for safer and more convenient weight loading.
- Original format:
.pth - Converted format:
.safetensors - Model architecture: U-Net++
- Backbone: EfficientNetV2
- Task: Manga text segmentation
The conversion changes only the weight serialization format; the model architecture and weights are otherwise intended to remain unchanged.
Original Model
Please refer to the original Hugging Face repository for the model description, dataset information, examples, and citation:
https://huggingface.co/a-b-c-x-y-z/Manga-Text-Segmentation-2025
Credits
Original model:
a-b-c-x-y-z — Manga-Text-Segmentation-2025
Original project:
Manga Text Segmentation — Segmentation of text in manga.
The original model card cites the work “Unconstrained Text Detection in Manga: A New Dataset and Baseline” by Julián del Gobbo and Rosana Matuk Herrera.
Please refer to the original repository for the applicable license and usage terms.
Files
.
├── model.safetensors
└── README.mdDisclaimer
This repository is a converted distribution of the original model weights. It is not the original model implementation.
For the original implementation, documentation, dataset information, and licensing details, please visit:
https://huggingface.co/a-b-c-x-y-z/Manga-Text-Segmentation-2025