The code snippet is below.
The ‘transforms’ argument can’t use?
The linked object detection tutorial links to the references/detection/transforms.py
, which implement transformations accepting an input and target tensor, so you would need to copy these fles (or git clone
) to your workspace.
Thank you, I get it.
But I want to know why not torchvision integrate these transforms(do same transform in image and target) internal?
I don’t know, but guess because there might have been concerns about overriding the default transformations.
I get it. Thank you.