Commit History
10 days ago
-
Add KanjiDNN v4: JIS Level 1+2 (6,541 classes), 98.6% on held-out real handwriting
Verified signatureb95bf73
-
Upload README.md with huggingface_hub
Verified signatureb1ae6fd -
Upload README.md with huggingface_hub
Verified signature570c442 -
Add KanjiDNN v3: PyTorch/ONNX model for finger-drawn kanji recognition
Verified signaturec9c52af
-
Upload config.json
Verified signature7619085
-
Rename KanjiDNN2_v2.keras to KanjiDNN_v2.keras
Verified signature84ca091 -
Rename KanjiDNN2.keras to KanjiDNN2_v2.keras
Verified signature1beade9 -
Upload KanjiDNN2.keras
Verified signatureaf251b3
-
Update README.md
Verified signature1d38e85 -
Update README.md
Verified signature86f7b7e -
Upload KanjiDNN_v1.keras
Verified signaturec325da7 -
Update README.md
Verified signature# 🧠 Model Description This is a Convolutional Neural Network (CNN) model designed to classify individual Japanese characters, including: - **Kanji** - **Kana** (Hiragana and Katakana) - **Romaji** (Latin characters) - **Punctuation** - **Digits** The model supports both **handwritten** and **printed** styles. It processes **64×64 grayscale** input images and outputs the corresponding character class. --- # 📚 Training Data The model was trained on a diverse set of sources: - **Self-generated images** - **SVG graphics** extracted from [KanjiVG](https://kanjivg.tagaini.net/) - **ETL9G dataset** from the [ETL Character Database](http://etlcdb.db.aist.go.jp/the-etl-character-database/) --- # 🔧 Intended Use This model is intended for: - Recognition of handwritten characters (similar to approaches like [sljfaq](https://kanji.sljfaq.org/)) - Serving as a basis for full-text OCR - Educational tools for character recognition - Research in handwriting recognition and multilingual OC…c80288e -
initial commit
Verified signaturee4ef7f8