cerebras.modelzoo.data.vision.segmentation.SkmDataProcessor.SkmDataProcessor#
- class cerebras.modelzoo.data.vision.segmentation.SkmDataProcessor.SkmDataProcessor(*args, **kwargs)[source]#
Bases:
cerebras.modelzoo.data.vision.segmentation.Hdf5BaseIterDataProcessor.Hdf5BaseIterDataProcessor
A SKM-TEA MRI DICOM Track (Stanford MRI Dataset) Data Processor class for U-Net Segmentation. This class includes data preprocessing and transforms that are necessary for utilizing the SkmDicomDataset class for training models.
Currently supports masks (segmentation) and does NOT support bounding boxes (detection).
- Parameters
params (dict) – YAML config file with adaptable model and data configurations
Methods
Classmethod to create the dataloader object.
preprocess_image
preprocess_mask
Preprocess the masks and images
- create_dataloader(is_training=True)#
Classmethod to create the dataloader object.