Publishing a YOLOv11x Model on Hugging Face

This article introduces the steps to publish a YOLOv11x model trained on the Japanese Classical Kuzushiji Dataset on Hugging Face and create a demo with Gradio Spaces.

Overview

• Model: YOLOv11x (for kuzushiji character detection)
• Dataset: Japanese Classical Kuzushiji Dataset
• Publication: Hugging Face Models + Spaces

1. Register the Model on Hugging Face Models

1.1 Install huggingface_hub

1.2 Login

Or from Python:

You can obtain a token from https://huggingface.co/settings/tokens (Write permission required).

1.3 Upload the Model

1.4 Create a Model Card (README.md)

Create a README.md with usage instructions and license information and upload it.

2. Publish a Demo on Hugging Face Spaces

2.1 Spaces Configuration (README.md)

Key points:

• Specifying sdk: gradio pre-installs Gradio
• sdk_version can be used to pin the version
• No need to add gradio to requirements.txt

2.2 requirements.txt

2.3 Load the Model in app.py

Use hf_hub_download to load the model from Hugging Face Hub.

2.4 Notes on Gradio 5.x

In Gradio 5.x, SSR (Server-Side Rendering) is enabled by default as an experimental feature. This can cause the following error:

To work around this issue, specify ssr_mode=False in launch().

3. Benefits of This Architecture

4. Published Resources

• Model: https://huggingface.co/nakamura196/yolov11x-codh-char
• Spaces: https://huggingface.co/spaces/nakamura196/yolov11x-codh-char

References

• Hugging Face Spaces - Gradio Documentation
• Ultralytics YOLO Documentation
• Japanese Classical Kuzushiji Dataset