GradCAM/CAM for multi-image input

[ltsc256]

I am using Monai for a classification task that uses both MR and CT data. The classification models work very well and I wanted to implement the interpretability methods (either occlusion sensitivity or cam). However, when I try to run cam, I get an error that forward is missing 1 required argument (the CT image) since the interpretability models seem to only allow 1 image input. Do you have any advice how to handle?

[KumoLiu]

Hi, @ltsc256, thanks for your interest here.
I would suggest you separate the inputs in the model. Something like:

def model(inputs):
      mri, ct = inputs["mri"], inputs["ct"]
      pass

Then you could directly send the inputs to cam/gradcam.

Hope it helps, thanks!