kha-white/manga-ocr
A specialized OCR model using Vision Encoder Decoder transformers to recognize Japanese text in manga images.
Not currently ranked — collecting fresh signals.
star history
This repository provides an optical character recognition system specifically optimized for Japanese manga. It uses a custom end-to-end model based on Hugging Face Transformers’ Vision Encoder Decoder architecture. The system handles manga-specific challenges including vertical and horizontal text orientation, furigana annotations, text overlaid on images, diverse font styles, and low-quality images. Unlike typical OCR tools, it processes multi-line text bubbles in a single forward pass without requiring line-by-line splitting.
Frequently asked
- What is kha-white/manga-ocr?
- A specialized OCR model using Vision Encoder Decoder transformers to recognize Japanese text in manga images.
- Is manga-ocr open source?
- Yes — kha-white/manga-ocr is open source, released under the Apache-2.0 license.
- What language is manga-ocr written in?
- kha-white/manga-ocr is primarily written in Python.
- How popular is manga-ocr?
- kha-white/manga-ocr has 2.7k stars on GitHub.
- Where can I find manga-ocr?
- kha-white/manga-ocr is on GitHub at https://github.com/kha-white/manga-ocr.