Hanson0910 / Pytorch-RIADD

1st solution for Retinal Image Analysis for multi-Disease Detection Challenge(RIADD (ISBI-2021))

Code for RIADD (ISBI-2021)

This is the source code for my solution to the RIADD (ISBI-2021) hosted by Grand Challenge

Requirement

• opencv_python==4.4.0.44
• tqdm==4.48.2
• pandas==1.1.1
• albumentations==0.4.6
• torchvision==0.6.0a0+35d732a
• torch==1.5.1
• visdom==0.1.8.9
• apex==0.9.10dev

Update Log

• [2021-2-28] Release the training codes
• [2021-4-14] publish final model weights

TODO

• publish partial model weights

Dataset

https://www.kaggle.com/sssdey1702/riadd-rfmid

Model Weights

https://www.kaggle.com/hanson0910/tf-efficientnet-b5-ns960

Train Model with DDP

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 train_riadd_eb6_ddp.py

Train Model without DDP

python train_riadd_eb6.py

Create submission

python subnmit_riadd.py

References

Appreciate the great work from the following repositories:

• https://github.com/rwightman/pytorch-image-models
• https://github.com/zlannnn