APTOS: data
IDRID: data
DeepDRiD: data
FGADR: data
Notes: you need to write emails to apply for the download link of this dataset
This paper, 'CANet: Cross-Disease Attention Network for Joint Diabetic Retinopathy and Diabetic Macular Edema Grading' could be followed.
This paper is about DR and DME jointly grading. You can only focus on DR grading and follow the format of the baseline.py in CANet.
Or you could follow the paper, 'Learning Robust Representation for Joint Grading of Ophthalmic Diseases via Adaptive Curriculum and Feature Disentanglement'.
Other papers available: 1. CABNet: Category Attention Block for Imbalanced Diabetic Retinopathy Grading
To run a deep learning algorithm by GPU, there are something need to be prepared.
link: https://developer.nvidia.com/cuda-downloads usually we download the newest version of cuda
link: https://pytorch.org/get-started/previous-versions/ if your gpu is old, something error may occur. you need to find the corresponding version of pytorch
open your terminal, type 'python' to get into python, then type like this:
the output should be true
link: https://zh.d2l.ai/chapter_convolutional-modern/resnet.html the d2l library contains some basic codes about the training process, how to compute the accuracy, how to visualize the output... you can check the codes for reference
link: https://github.com/tensorflow/tensorboard
- a dataloader to load your DR dataset
- use parsers to imput as commands
- load your model (usually Resnet) and use pretrained model
- the training process
- compute the accuracy and auc score
- visualize the output
