basveeling / keras-gcnn

Roto-reflection equivariant CNNs for Keras as presented in B. S. Veeling, J. Linmans, J. Winkens, T. Cohen, M. Welling. "Rotation Equivariant CNNs for Digital Pathology".
Other
85 stars 42 forks source link

bugfix; makes GBatchNorm centering and scaling optional #7

Closed neel-dey closed 3 years ago

neel-dey commented 4 years ago

Hi Bas,

This PR lets GBatchNorm make the affine transformation ( gamma * x + beta ) at the end optional.

This functionality is present in the current version of the function but the code as is tries to apply the repeat() function on None, leading to a crash.

Thanks for the great library for equivariant learning!

Best, Neel