[Question] How to pass the batch size wenn fine tuning the SAM model on own dataset?

[wilhem]

I have a problem, when passing more than one element to the network during the training phase on my dataset.

The problem is here:

predictor.set_image(image)

if the image has shape: (16, 1024, 1024, 3) where 16 is the batch size, and 3 the number of channels,
then I get an error, that the dimension does not match the desired input.
If the batch size is 1, then it works.

How to deal with it?