Is your feature request related to a problem? Please describe.
There some use cases for changing the BatchNorm layers in the backbones, especially when using transfer learning it would be more efficient to use a frozen batchnorm in the backbone.
Describe the solution you'd like
support changing BatchNorm layer in all backbones, using norm_layer argument like those in resnet torchvision
Is your feature request related to a problem? Please describe. There some use cases for changing the BatchNorm layers in the backbones, especially when using transfer learning it would be more efficient to use a frozen batchnorm in the backbone.
Describe the solution you'd like support changing BatchNorm layer in all backbones, using
norm_layerargument like those in resnet torchvision