Closed anuragp2018 closed 2 months ago
No preprocessing is needed. But your image seems quite different to the training data, so bad results are not surprising. See On the Compactness, Efficiency, and Representation of 3D Convolutional Networks: Brain Parcellation as a Pretext Task for more information.
Is any preprocessing required before running the module? I tried it on T1-weighted MRI, but it gave a different output compared to your sample data output.