Choom

A project for training and testing Chinese character recognition using CNN models and GNT/MPF dataset parsing.

Features

• Convert .gnt files to PNG visualizations (with GB2312 data and simplified Chinese character data)
• Clean, readable CNN architecture for classification
• Data loaders for custom dataset ingestion
• Utilities for experimenting with different model architectures

Installation

pip install -r requirements.txt

Dataset

Choom supports the handwriting datasets stored in .gnt format. The project includes dedicated utilities to parse these files and export them to images.

Usage

python infer.py --image your_image.png

License

MIT License