This repository contains the source code for developing a deep learning system in detecting corneal (DC) diseases from low-quality slit lamp images The performance of a deep learning system in detecting corneal diseases from low-quality images can be improved if low-quality images with sufficient diagnostic certainty are added to the training set.

• Ubuntu: 18.04 lts
• Python 3.7.8
• Pytorch 1.6.0
• NVIDIA GPU + CUDA_10.0 CuDNN_7.5 This repository has been tested on NVIDIA RTX2080Ti. Configurations (e.g batch size, image patch size) may need to be changed on different platforms.

Other packages are as follows:

• wheel
• yaml
• scipy
• joblib
• opencv-python
• scikit-image
• numpy

• pip install -r requirements.txt

|-data  
    |-train  
       |--label1  
           |--*.jpg  
       |--label2  
           |--*.jpg  
       |--label3      
           |--*.jpg  
    |-val   
     ...  
    |-test   
     ...

The training and testing are executed as follows:

• The file "CDtrainingv1.py" in /CD-system-Source is used for our models training.

python3 CDtrainingv1.py --data PATH

• The file "CDtestingv1.py" in /CD-system-Source is used for testing.

python3 CDtestingv1.py

Please feel free to contact us for any questions or comments: Zhongwen Li, E-mail: [email protected] or Jiewei Jiang, E-mail: [email protected].