[Enhancement] Support padding in testing #2290

xiexinch · 2022-11-10T06:43:09Z

Motivation

Fix removing gt_sem_seg padding at postprocess_result
Support different padding methods in testing

Modification

Add removing gt_sem_seg padding at postprocess_result method in mmseg/models/segmentors/base.py.
Add test_cfg` params to SegDataPreprocessor, supporting different padding methods.

BC-breaking (Optional)

None

Use cases

In a case that uses size in training and using size_divisor in testing, we might set the test_cfg param to data_preprocessor.

data_preprocessor = dict(
    type='SegDataPreProcessor',
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    bgr_to_rgb=True,
    pad_val=0,
    seg_pad_val=255,
    size=(512, 512),
    test_cfg=dict(size_divisor=32))

MeowZheng

add ut for the new function

MeowZheng · 2022-11-10T09:14:34Z

mmseg/models/data_preprocessor.py

+ train_cfg (dict, optional): The padding size config in training, if
+ not specify, will use `size` and `size_divisor` params as default.
+ Defaults to None, only supports keys `size` or `size_divisor`.
+ test_cfg (dict, optional): The padding size config in testing, if not
+ specify, will use `size` and `size_divisor` params as default.
+ Defaults to None, only supports keys `size` or `size_divisor`.


Please add an example in PR description for train_cfg and test_cfg as a padding config for input data.

codecov · 2022-11-10T10:54:59Z

Codecov Report

Base: 83.57% // Head: 83.65% // Increases project coverage by +0.08% 🎉

Coverage data is based on head (8fed7f5) compared to base (38c3145).
Patch coverage: 100.00% of modified lines in pull request are covered.

❗ Current head 8fed7f5 differs from pull request most recent head 4e12bcd. Consider uploading reports for the commit 4e12bcd to get more accurate results

Additional details and impacted files

@@             Coverage Diff             @@
##           dev-1.x    #2290      +/-   ##
===========================================
+ Coverage    83.57%   83.65%   +0.08%     
===========================================
  Files          141      141              
  Lines         7970     7974       +4     
  Branches      1192     1193       +1     
===========================================
+ Hits          6661     6671      +10     
+ Misses        1119     1114       -5     
+ Partials       190      189       -1

Flag	Coverage Δ
unittests	`83.65% <100.00%> (+0.08%)`	⬆️

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

Impacted Files	Coverage Δ
mmseg/models/data_preprocessor.py	`93.02% <100.00%> (+0.52%)`	⬆️
mmseg/models/segmentors/base.py	`87.50% <100.00%> (+0.19%)`	⬆️
mmseg/utils/misc.py	`90.90% <0.00%> (+13.63%)`	⬆️

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.

MeowZheng

please try to use Pad in test pipeline

xiexinch · 2022-11-16T09:16:38Z

please try to use Pad in test pipeline

If use Pad in test _pipeline, the padded pixels would be normalized.
https://github.com/open-mmlab/mmsegmentation/blob/dev-1.x/mmseg/models/data_preprocessor.py#L108

MeowZheng · 2022-11-19T08:07:31Z

mmseg/models/data_preprocessor.py

+ inputs, data_samples = stack_batch(
+ inputs=inputs,
+ data_samples=data_samples,
+ size=self.test_cfg.get('size', None),
+ size_divisor=self.test_cfg.get('size_divisor', None),
+ pad_val=self.pad_val,
+ seg_pad_val=self.seg_pad_val)


Suggested change

inputs, data_samples = stack_batch(

inputs=inputs,

data_samples=data_samples,

size=self.test_cfg.get('size', None),

size_divisor=self.test_cfg.get('size_divisor', None),

pad_val=self.pad_val,

seg_pad_val=self.seg_pad_val)

inputs, data_samples = stack_batch(

inputs=inputs,

size=self.test_cfg.get('size', None),

size_divisor=self.test_cfg.get('size_divisor', None),

pad_val=self.pad_val,

seg_pad_val=self.seg_pad_val)

I think it would be better to keep gt as original when model testing or validation.

MeowZheng · 2022-11-19T08:07:57Z

mmseg/models/segmentors/base.py

+ i_gt_sem_seg = data_samples[i].gt_sem_seg.data[:,
+ padding_top:H -
+ padding_bottom,
+ padding_left:W -
+ padding_right]
+


Suggested change

i_gt_sem_seg = data_samples[i].gt_sem_seg.data[:,

padding_top:H -

padding_bottom,

padding_left:W -

padding_right]

MeowZheng · 2022-11-19T08:09:22Z

mmseg/models/segmentors/base.py

+ 'gt_sem_seg':
+ PixelData() if only_prediction else PixelData(


Is it necessary to add an empty PixelData for gt_sem_seg when only predict?

* support padding in test and fix remove gt padding at post_process * fix get gt data * fix ut * add data_preprocessor ut * remove gt padding * fix data sample is None

Dev 1.x

[Enhancement] Support padding in testing

xiexinch added 2 commits November 10, 2022 14:21

support padding in test and fix remove gt padding at post_process

70daaaa

fix get gt data

d0b35cd

mm-assistant bot assigned MengzhangLI Nov 10, 2022

xiexinch added the 1.x Related issue of 1.x version label Nov 10, 2022

MeowZheng reviewed Nov 10, 2022

View reviewed changes

fix ut

d33af52

MeowZheng reviewed Nov 11, 2022

View reviewed changes

MeowZheng added this to the 1.0.0rc2 milestone Nov 11, 2022

add data_preprocessor ut

8fed7f5

xiexinch changed the title ~~[Enhancement] Support different padding methods in training and testing~~ [Enhancement] Support different padding in testing Nov 17, 2022

xiexinch changed the title ~~[Enhancement] Support different padding in testing~~ [Enhancement] Support padding in testing Nov 17, 2022

MeowZheng approved these changes Nov 19, 2022

View reviewed changes

MeowZheng reviewed Nov 19, 2022

View reviewed changes

xiexinch added 2 commits November 19, 2022 17:38

remove gt padding

f915df4

fix data sample is None

4e12bcd

MeowZheng approved these changes Nov 19, 2022

View reviewed changes

MeowZheng merged commit c56a299 into open-mmlab:dev-1.x Nov 19, 2022

aravind-h-v pushed a commit to aravind-h-v/mmsegmentation that referenced this pull request Mar 27, 2023

fix pix2pix docs (open-mmlab#2290)

648090e

wjkim81 pushed a commit to wjkim81/mmsegmentation that referenced this pull request Dec 3, 2023

Merge pull request open-mmlab#2290 from open-mmlab/dev-1.x

98ccc8a

Dev 1.x

nahidnazifi87 pushed a commit to nahidnazifi87/mmsegmentation_playground that referenced this pull request Apr 5, 2024

Merge pull request open-mmlab#2290 from xiexinch/fix_gt_padding

289716e

[Enhancement] Support padding in testing

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Enhancement] Support padding in testing #2290

[Enhancement] Support padding in testing #2290

xiexinch commented Nov 10, 2022 •

edited

Loading

MeowZheng left a comment

MeowZheng Nov 10, 2022

codecov bot commented Nov 10, 2022 •

edited

Loading

MeowZheng left a comment

xiexinch commented Nov 16, 2022

MeowZheng Nov 19, 2022 •

edited

Loading

MeowZheng Nov 19, 2022

MeowZheng Nov 19, 2022

xiexinch Nov 19, 2022

[Enhancement] Support padding in testing #2290

[Enhancement] Support padding in testing #2290

Conversation

xiexinch commented Nov 10, 2022 • edited Loading

Motivation

Modification

BC-breaking (Optional)

Use cases

MeowZheng left a comment

Choose a reason for hiding this comment

MeowZheng Nov 10, 2022

Choose a reason for hiding this comment

codecov bot commented Nov 10, 2022 • edited Loading

Codecov Report

MeowZheng left a comment

Choose a reason for hiding this comment

xiexinch commented Nov 16, 2022

MeowZheng Nov 19, 2022 • edited Loading

Choose a reason for hiding this comment

MeowZheng Nov 19, 2022

Choose a reason for hiding this comment

MeowZheng Nov 19, 2022

Choose a reason for hiding this comment

xiexinch Nov 19, 2022

Choose a reason for hiding this comment

xiexinch commented Nov 10, 2022 •

edited

Loading

codecov bot commented Nov 10, 2022 •

edited

Loading

MeowZheng Nov 19, 2022 •

edited

Loading