running the image upscale for the various upscaling algorithms
Upscale Community 不同场景下的增强画质模型,支持不同的商业用途!
一些模型的size 比较大,上传会显示失败!在部署前的预处理先下载这些模型,并存放指定路径。
人脸超分模型:
检测人脸模型:
detection_Resnet50_Final.pth
parsing_parsenet.pth
Image-Upscale/
|
|-- assets/ # Directory to store assets
|-- bin/ # Directory for image-upscale executor
|-- master/ # Directory for main processors
| |-- figure # Preprocessing functions and utilities
| |-- |-- figureInference.py # Figure model processing scripts
| |-- gfpgan # Detect figure basic model
| |-- |-- weights
| |-- |-- |-- detection_Resnet50_Final.pth # Detect protrait
| |-- |-- |-- parsing_parsenet.pth # Parsing protrait
| |-- constants.py # File contains all default constants
| |-- instance.py # Input args test file
| |-- logsConfig.py # Logs config file
| |-- main.py # main file
|
|-- models/ # Directory for saving all models
|
|-- output/ # Directory for output image file
|-- .gitignore # not uploaded file or dir
|-- LICENSE # MIT LICENSE
|-- README.md # Project documentation and instructions
|-- requirements.txt # List of dependencies and required libraries
master/constants.py的 DEBUG_MODE 决定日志是否开启debug模式
@article{wang2022restoreformer,
title={RestoreFormer: High-Quality Blind Face Restoration from Undegraded Key-Value Pairs},
author={Wang, Zhouxia and Zhang, Jiawei and Chen, Runjian and Wang, Wenping and Luo, Ping},
booktitle={The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2022}
}