L1 loss between a single image and batch of images

1

Hi,
I’m trying to assess which loss to use in some project and I need to calculate loss between a single image and batch of images.

So far this is what I have is this:

Photo from Sagy
Photo from Sagy805×191 34.8 KB

Which seems unnecessary since I don’t really need the same images to appear batch_num times.

Any other way? if i give only one image instead of copying it to be the same size? will by broadcasting it will achieve the same result?

2

You can use expand method to the single image. It points to the same mem adress than the original image.
https://pytorch.org/docs/stable/tensors.html#torch.Tensor.expand

3

Thank didn’t know that existed, good idea