- Lung nodule segmentation implemented in Pytorch.
- Read and implement 2 papers:
- Python >= 3.5 (3.6 recommended).
pip install -r requirements.txt
.
├── backbone
│ ├── __init__.py
│ ├── modules.py
│ ├── res_unet_plus.py
│ ├── res_unet.py
│ └── unet.py
├── checkpoints
│ └── default
│ └── exp8_512.pt - pretrained model with image size 512 x 512
├── config
│ └── default.yaml - holds configuration for training
├── dataset
│ └── dataloader.py - anything about data loading goes here
├── inference.py
├── luna_mask_extraction.py - extracts raw data to image, mask
├── lung_segment_png.py
├── preprocess.py
├── README.md
├── requirements.txt
├── train.py
└── utils
├── augmentation.py
├── hparams.py
├── __init__.py
├── logger.py
└── metrics.py
6 directories, 20 files
- Training
- From scratch:
python train.py --name "default" --config "config/default.yaml"
- Load from:
python train.py --name "default" --config "config/default.yaml" \ --load_from "checkpoints/default/exp8_512.pt"
- From scratch:
- Inference:
python inference.py