Repo containing code where SAM was fine-tuned on IWOAI 2019 knee MRI data. A 3D U-Net was also trained as a baseline.
Models, dataset classes, metrics and utility functions are all defined in src.
Train scripts for models are found in scripts folder.
notebooks contains all jupypter notebooks, most of which just test the code contained in the train scripts.
Other notebooks are for extracting results from the generated masks (Bland_Altman_Plots.ipynb, Dice_scores.ipynb, Hausdorff_Distance.ipynb)
data folder is empty. Put data here after cloning.
Directory Tree:
.
├── LICENSE
├── README.md
├── data
│ └── data.md
├── knees.yml
├── models
│ └── models.md
├── notebooks
│ ├── Bland_Altman_Plots.ipynb
│ ├── Dice_scores.ipynb
│ ├── Hausdorff_Distance.ipynb
│ ├── convert_to_slices.ipynb
│ ├── run_test_split_through_sam.ipynb
│ ├── test_sam.ipynb
│ ├── test_sam_slice_files.ipynb
│ └── test_unet.ipynb
├── scripts
│ ├── hyperparams_sam.txt
│ ├── hyperparams_unet.txt
│ ├── train_SAM.py
│ ├── train_SAM_slices.py
│ └── train_UNet.py
└── src
├── datasets.py
├── metrics.py
├── model_SAM.py
├── model_UNet.py
└── utils.py