crop_size and img_scale order

In base dataset-related configs you usually have something like, e.g. in cityscapes

crop_size = (512, 1024)
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations'),
    dict(type='Resize', img_scale=(2048, 1024), ratio_range=(0.5, 2.0)),

does it mean that crop_size has height-width order while img_scale in Resize has width-height order as cityscapes is obviously has the width equal to ~2x of the height and crops are expected to have similar ratio?

[xvjiarui]

Hi @nizhib For term size, we usually use hight-width order in this codebase. For term img_scale in Resize transform, the order doesn’t matter. The smaller one will be considered as the short side, the larger one will be considered as the long side. Specifically, img_scale=(2048, 1024) means the max scale of the resized image is of long side 2048, short side 1024.

[ParashRahman]

After investigating, I found that mmcv.imresize takes in (w,h) while mmcv.pad takes in (h,w). When img_scale=(w,h) is passed as a tuple to mmseg with no other scaling parameters, it is passed to mmcv.imresize(w,h) properly.

The way mmseg uses mmcv needs crop and pad to be (h, w).