🈶 Japanese Character Recognition CNN (Keras)
A Convolutional Neural Network (CNN) model built with Keras to classify individual Japanese characters from 64×64 grayscale images. Supports both handwritten and printed styles.
🧠 Character Classes Supported
- Kanji
- Hiragana and Katakana (Kana)
- Romaji (Latin letters)
- Digits (0–9)
- Common punctuation marks
📚 Training Data
The model was trained on a diverse dataset combining:
- 🧪 Self-generated character images
- 🖋️ SVG graphics from KanjiVG
- 🗃️ ETL9G dataset from the ETL Character Database
🚀 Try the Model in Action
You can interact with this model live via Gradio demo in Hugging Face Space.
- Downloads last month
- 19