Input mask prompt

[theFilipko]

Hello :)
I would like to ask you why do you resize the label used as a prompt to 256x256 without padding?
https://github.com/SysCV/sam-hq/blob/ac19724c47b13689e5d9596277a6522b371001c8/train/train.py#L412C40-L412C46

My understanding is, that for SAM you need to resize the longest side of the image (1024px), also the input prompts. SAM internally makes the input image square by padding with top-left alignment. The points and boxes SAM does not adjust because of the top-left alignment. Therefore, I suppose, that the same should be applied to the input mask (prompt), otherwise the prompt is not aligned with the image.