manga ocr

Introduction

MANGA OCR is an optical character recognition (OCR) tool designed for recognizing Japanese text, primarily focusing on Japanese manga. It is robust against various scenarios unique to manga, such as handling both vertical and horizontal text, text with furigana, text overlaid on images, diverse fonts and styles, and low-quality images.

Architecture

This OCR tool leverages the Vision Encoder Decoder framework, which is a part of the Hugging Face Transformers library. This architecture allows it to effectively convert images with Japanese text into textual data.

Training

MANGA OCR is trained on the manga109s dataset, which includes a substantial collection of Japanese manga, providing the model with a wide array of text scenarios typical in manga content.

Guide: Running Locally

1. Setup Environment: Ensure you have Python and Git installed on your system.
2. Clone Repository: Clone the MANGA OCR repository from GitHub.
3. Install Dependencies: Navigate to the cloned directory and install necessary dependencies using pip install -r requirements.txt.
4. Run OCR: Execute the OCR script on your image files containing Japanese text.

For better performance, consider using cloud GPUs such as those provided by AWS, Google Cloud, or Azure.

License

MANGA OCR is licensed under the Apache-2.0 License, which allows for permissive use, distribution, and modification.