How to apply Group Normalization to PyTorch FasterRCNN training?

I am following this implementation of RCNN (I will use FasterRCNN instead of MaskRCNN)

I want to use Group Normalization for the input layer in my DataSet class. I don’t want to use custom pipeline and create my DataSet and DataLoader since PyTorch training pipeline is very convenient to use.

Is it possible to use GroupNormalization with standard Dataloader and Dataset from

There is one (I think) available DataLoader: -> and couple DataSets: -> Dataset: -> IterableDataset: