Descriptions of scripts in this repository are listed below.
- train_mrcnn.py - file to train a Mask R-CNN algorithm to segment the CA and IJV from neck US images. The data used to train can be found under the PARTITION FOR CROSS-VALIDATION, this provides the start and end locations for a four-fold cross validation.
- train_unet.py - file to train a U-Net algorithm to segment the CA and IJV from neck US images. The data used to train can be found under the PARTITION FOR CROSS-VALIDATION, this provides the start and end locations for a four-fold cross validation.
- test_mrcnn.py - Test Mask R-CNN performance, display predictions of the Mask R-CNN model
- test_unet.py - Test U-Net performance, display predictions of the U-Net model
- unet.py - U-Net model implementation
- mrcnnconfig.py - Mask R-CNN base Config class
- mrcnnmodel.py - Mask R-CNN model implementation
- mrcnnparrallel_model.py - Mask R-CNN Multi-GPU support
- mrcnnsubclass.py - Subclasses of the Mask R-CNN Config and Dataset classes
- mrcnnutils.py - Mask R-CNN common utility functions and classes.
- mrcnnvisualize.py - Mask R-CNN display and visualization functions.
Please go to https://1drv.ms/u/s!Akxm4gUER2IFag1CJ1LtM_qB6HY?e=0aqATl to download the numpy image arrays.
We express our gratitude to Matterport, as we extended their Mask R-CNN implementation to make this project possible.