A django-based website for displaying Diabetic Retinopathy Auto-Detection.
The website requires django being installed.
- It aims to enable users to upload their own fundus photos to be classified by CNN models.
- However, this website does not link to any CNN.
- Data in server has been generated by the first place team of Kaggle Diabetic Retinopathy competition
- Download the website directory.
- Install django by opening the terminal and entering the following command:
pip install django
- Change to the website directory.
- Execute:
python manage.py runserver
And you will see the website in browser with localhost:8000/diabetics/ (by default)
Directory | Content |
---|---|
Model | Three trained CNN model files. |
Data | Preprocessed 300x300 sample images.(Unpreprocessed are in deepsite/media) |
Script | Preprocessing and model-averaging scripts. |
CNN | Modified cnn codes for small sample testing. |
- Fork SparseConvNet, and switch to kaggle branch.
git clone https://github.com/btgraham/SparseConvNet.git
cd SparseConvNet
git checkout kaggle_Diabetic_Retinopathy_competition
- Replace files with our modified files:
- Put files in Model into "kaggleDiabeticsRetinopathyModelfiles".
- Put files in website/media into "Data/kaggleDiabeticRetinopathy/sample/".
- Put files in Script into "Data/kaggleDiabeticRetinopathy/".
- Put codes in CNN into SparseConvNet directory.
- In "Data/kaggleDiabeticRetinopathy/", execute:
./createSample.sh
python preprocessImages.py
- In SparseConvNet/, execute:
make kaggleDiabetes1
./kaggleDiabetes1
- Then the classification result will be stored as csv file.
- Upload multiple images.
- View uploaded images in table.
- Simple barchart of numbers of severities.
- Individual inspection of each uploaded images.
- Real-time classification with CNN models.