Stacking different input types with FastAI

I’m trying to figure out how to stack different input methods together, for different output.
In FastAi, we have the PointBlock. But can we combine input such that we have x input as images and points, and y output as images and points. I want to do regression on the input points, using image as additional input data. @sgugger would you have any ideas I can follow?

@muellerzr I know you’ve done work on this area so would you have any notebooks like this?
Maybe it is already there in DataBlock api and I am missing it?

1 Like