Pre-Trained Models for the Uformer Hub Service

This repository hosts the official, pre-trained model weights for the Uformer: A General U-Shaped Transformer for Image Restoration paper.

How to Use These Models

❗ Important: These model files are intended to be used with the Uformer Hub Service, a full-stack, containerized application that makes running these models simple.

➡️ Get the application here: https://github.com/CaveMindLabs/uformer-hub-service

The main application provides a user-friendly web interface, an API, and all the necessary code to run these models for image, video, and live-stream enhancement. Please follow the setup instructions in the main repository's README.md file.

The instructions will guide you to download the necessary files from this repository and place them in the correct directory (backend/model_weights/official_pretrained/).

Models Included

This repository contains the following pre-trained models:

Model File	Task	Training Dataset
`Uformer_B_SIDD.pth`	Denoising (High Quality)	SIDD
`uformer16_denoising_sidd.pth`	Denoising (Fast)	SIDD
`Uformer_B_GoPro.pth`	Motion Deblurring	GoPro

License

The model weights are released under the original project's MIT License.

Citation & Acknowledgements

Full credit for these models goes to the original authors of the Uformer paper.

@InProceedings{Wang_2022_CVPR,
    author    = {Wang, Zhendong and Cun, Xiaodong and Bao, Jianmin and Zhou, Wengang and Liu, Jianzhuang and Li, Houqiang},
    title     = {Uformer: A General U-Shaped Transformer for Image Restoration},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {17683-17693}
}

Original Codebase: github.com/ZhendongWang6/Uformer