Although in their purest form the ultrasound images should be single-channel grayscale images, some datasets are saved as 3-channel images with non-identical channels. For consistency, I propose outputting the remain datasets as 3-channel images by replicating the single-channel image in the channel dimension. The benefit of this approach is that the same model architecture can be used without requiring dataset-specific preprocessing.
Although in their purest form the ultrasound images should be single-channel grayscale images, some datasets are saved as 3-channel images with non-identical channels. For consistency, I propose outputting the remain datasets as 3-channel images by replicating the single-channel image in the channel dimension. The benefit of this approach is that the same model architecture can be used without requiring dataset-specific preprocessing.