AIbase
EN
Home
Kanjidnn
K

Kanjidnn

Developed by gaiseras
A convolutional neural network (CNN) built on the Keras framework, specifically designed to recognize individual Japanese characters from 64ร—64 grayscale images, supporting both handwritten and printed text recognition.
Text Recognition JapaneseOpen Source License:Apache-2.0#Japanese mixed character recognition#Handwritten and printed dual modality#64x64 grayscale image adaptation
Downloads 38
Release Time : 5/12/2025

Model Overview

This model can recognize various Japanese characters, including kanji, hiragana, katakana, Roman letters, numbers, and common punctuation marks, making it suitable for image-to-text tasks.

Model Features

Multi-source training data
The model is trained using self-generated character images, SVG vector graphics from KanjiVG, and the ETL9G dataset from the ETL character database to ensure recognition accuracy.
Supports multiple character types
Capable of recognizing kanji, hiragana, katakana, Roman letters, numbers, and common punctuation marks, covering a wide range of Japanese characters.
Handwritten and printed text compatibility
The model supports recognition of both handwritten and printed Japanese characters, adapting to different application scenarios.

Model Capabilities

Japanese character recognition
Image-to-text
Handwritten text recognition
Printed text recognition

Use Cases

Document digitization
Japanese document scanning and recognition
Convert scanned Japanese documents into editable text formats.
Improves document processing efficiency and reduces manual input errors.
Education
Japanese learning assistance
Recognize handwritten Japanese characters by students and provide instant feedback.
Helps learners quickly master Japanese character writing.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English็ฎ€ไฝ“ไธญๆ–‡็น้ซ”ไธญๆ–‡ใซใปใ‚“ใ”
ยฉ 2025AIbase