[Fix] Fix ResizeToMultiple transform in MMSeg 1.x

[MengzhangLI]

Motivation

Fix ResizeToMultiple transform in MMSeg 1.x.

Description

In current ResizeToMultiple, it has:

        results['img'] = img
        results['img_shape'] = img.shape
        results['pad_shape'] = img.shape

If dict(type='ResizeToMultiple', size_divisor=32), is added, because in https://github.com/open-mmlab/mmsegmentation/blob/dev-1.x/mmseg/models/utils/wrappers.py#L27, the input and size should have same dimension, but currently it is 2 and 3, respectively. And it would cause error ValueError: size shape must match input shape. Input is 2D, size is 3:

Solution

Just modify ResizeToMultiple:

        results['img'] = img
        results['img_shape'] = img.shape[:2]
        results['pad_shape'] = img.shape[:2]

[codecov]

Codecov Report

Base: 83.53% // Head: 83.55% // Increases project coverage by +0.02% 🎉

Coverage data is based on head (7d4ee84) compared to base (8b39294).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@             Coverage Diff             @@
##           dev-1.x    #2185      +/-   ##
===========================================
+ Coverage    83.53%   83.55%   +0.02%     
===========================================
  Files          138      138              
  Lines         7790     7790              
  Branches      1167     1167              
===========================================
+ Hits          6507     6509       +2     
+ Misses        1106     1104       -2     
  Partials       177      177              

Flag	Coverage Δ
unittests	83.55% <100.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmseg/datasets/transforms/transforms.py	96.93% <100.00%> (ø)
mmseg/models/backbones/timm_backbone.py	45.83% <0.00%> (+8.33%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[ThorPham]

I has same error like this when i train customer dataset. But i can't fix with modify
results['img'] = img results['img_shape'] = img.shape[:2] results['pad_shape'] = img.shape[:2]
I printed function resize in wrappers.py
I don't know why the last is shape Size([512, 512, 3])

Here is log error