OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Yuzumarker.fontdetection

Yuzumarker.fontdetection is a Python computer-vision model and implementation for recognizing Chinese, Japanese and Korean fonts in scene-text images and extracting font styles. The repository also includes scripts for generating and checking scene-text font datasets, and links to an online demo hosted on Hugging Face Spaces.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Yuzumarker.fontdetection — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use the online demo to explore the recognition task, or use the repository when you need to study the implementation or generate scene-text training data with your own background images.

Who it is for

  • People exploring CJK font recognition through the linked online demo.
  • Developers or researchers building scene-text datasets for font-recognition work.
  • Users adapting or studying a Python and PyTorch computer-vision implementation.

Documented uses

  • Recognize CJK fonts in rendered scene-text images.
  • Extract font styles from CJK scene text.
  • Generate and check a dataset of scene-text images rendered with different fonts.
Before you choose

Requirements & limits

The repository is public and lists an MIT license. The README links to a Hugging Face Spaces online demo.

  • The project targets Chinese, Japanese and Korean font recognition and style extraction.
  • The online demo is a separate entry point from the repository's dataset-generation scripts.
  • Dataset generation requires a CJK font pack and user-prepared background images; the original Pixiv backgrounds are not shared.
  • The README documents a Linux virtual-environment setup requiring python3-venv and Python 3.
  • Technical metadata identifies Python, PyTorch, Docker and MIT licensing.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Generate a scene-text font dataset

Prepare inputs, generate rendered CJK scene-text examples, and check generated files.

Before you start

  • A CJK font pack extracted into dataset/fonts.
  • Background images placed in dataset/pixivimages; the project does not share the original background images.
  • Python available to run the documented scripts.

Steps

  1. Run the command below to clean file names.

    Documented command
    python dataset_filename_preprocess.py

  2. Run the command below to generate a dataset in one partition, or use the documented two-parameter partition pattern for parallel tasks.

    Documented command
    python font_ds_generate_script.py 1 1

  3. Run the command below to check for corrupted generated images and labels; rerun generation if needed to fill removed files.

    Documented command
    python font_ds_detect_broken.py

Expected outcome

Generated dataset files are saved in dataset/font_img.

Guide limitations

  • The README notes the cache may not detect corruption if generation is interrupted while writing files.
  • The project's original Pixiv background images are not shared; users must prepare their own.
Documentation 1 ↗
Source-based practical guide

Set up the documented Linux generation environment

Create the project's virtual environment for dataset-generation scripts on a Linux cluster.

Before you start

  • A Linux cluster with python3-venv installed and python3 available in PATH.

Steps

  1. Run ./linux_venv_setup.sh.

  2. The script creates a virtual environment in venv and installs required packages, including libraqm as described in the README.

Expected outcome

A virtual environment is created for the documented dataset-generation workflow.

Guide limitations

  • This covers environment setup, not scheduling or running a generation task.
  • The README notes libraqm is needed for PIL text rendering and is often absent by default on Linux server distributions.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
JeffersonQin/YuzuMarker.FontDetection · Official homepage title · retained in its source language
How we review sourcesSuggest a correction