Pixel Base Image Quality

This deep neural network model detect the input image (left) and return its quality (right). Brightness represents quality, where brighter areas indicate higher quality and darker areas indicate lower quality.

---

The inference codes for the image quality detection based on the pixel. It takes as input a image and return a quality feature map. Training data generation and model training are done by our team. Model code adapted from: https://github.com/HuCaoFighting/Swin-Unet. citation as follows:[Cao et al., ECCVW 2022 / arXiv:2105.05537].

---

1. Download pre-trained model

• [Get pre-trained model in this link] (https://drive.google.com/drive/folders/1GqToFXaKdNUus3Cgp8Dtw2IRJrxSdNUA?usp=sharing): Put pretrained model into folder "model/"

2. Environment

• Prepare an environment with python=3.9, and then use the command "pip install -r requirements.txt" for the dependencies.

3. Run the model

Run the code by following command. For example:

python image_quality_command.py -i "C:\Users\User\Desktop\report_image" -o "C:\Users\User\Desktop\output"

Specify input data folder using -i and output folder using -o.

4. References

• SwinUnet
• SwinTransformer