U-Net++ model for SIIM-ACR-Pneumothorax-Seg-XR dataset. For original repo, see UNet++
SIIM-ACR-Pneumothorax-Seg-XR dataset.
python = 3.7, PyTorch = 1.7.1, cuda = 10.1
1*Nvidia RTX 3090 24GB
docker pull stevezeyuzhang/colab:1.7.1
pip install -r requirements.txt
|-- U-Net-_SIIM
|-- inputs
|-- SIIM-ACR-Pneumothorax-Seg-XR
|-- images
|-- <your image>
|-- masks
|-- 0
|-- <your label>(the same name with image)
|-- SIIM-ACR-Pneumothorax-Seg-XR_test (this is for segmentation)
|-- images
|-- <your image>
python train.py --dataset SIIM-ACR-Pneumothorax-Seg-XR --arch NestedUNet --img_ext .png --mask_ext .png
The checkpoint is saved in models
python inference.py --name SIIM-ACR-Pneumothorax-Seg-XR_NestedUNet_woDS
The results will be in the outputs .
You may need to resize the the ouputs to size corresponding with the test images. See inference.py.