Skip to content

[Fix] Support setting pad_val in batch collating#6837

Merged
ZwwWayne merged 2 commits intoopen-mmlab:dev-v2.20.0from
AronLin:pad
Dec 24, 2021
Merged

[Fix] Support setting pad_val in batch collating#6837
ZwwWayne merged 2 commits intoopen-mmlab:dev-v2.20.0from
AronLin:pad

Conversation

@AronLin
Copy link
Contributor

@AronLin AronLin commented Dec 20, 2021

Now, the padding value is set to 0 by default, but in gt_semantic_seg it needs to be 255.

So I modified the DefaultFormatBundle and added padding_value in the DataContainer.
In batch collating, the DataContainer will fill the data with the padding_value.

@AronLin
Copy link
Contributor Author

AronLin commented Dec 20, 2021

I will run an HTC model and a Panoptic FPN model to ensure performance stability.

@codecov
Copy link

codecov bot commented Dec 20, 2021

Codecov Report

Merging #6837 (348ce0e) into dev-v2.20.0 (5e246d5) will increase coverage by 0.01%.
The diff coverage is 100.00%.

Impacted file tree graph

@@               Coverage Diff               @@
##           dev-v2.20.0    #6837      +/-   ##
===============================================
+ Coverage        62.02%   62.03%   +0.01%     
===============================================
  Files              322      323       +1     
  Lines            25717    25745      +28     
  Branches          4356     4363       +7     
===============================================
+ Hits             15951    15971      +20     
- Misses            8918     8928      +10     
+ Partials           848      846       -2     
Flag Coverage Δ
unittests 62.01% <100.00%> (+0.01%) ⬆️

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

Impacted Files Coverage Δ
mmdet/datasets/pipelines/formating.py 67.76% <100.00%> (+0.26%) ⬆️
mmdet/models/detectors/cornernet.py 94.87% <0.00%> (-5.13%) ⬇️
mmdet/models/dense_heads/corner_head.py 68.06% <0.00%> (-1.41%) ⬇️
mmdet/apis/train.py 15.66% <0.00%> (-1.01%) ⬇️
mmdet/datasets/lvis.py 14.86% <0.00%> (ø)
mmdet/utils/__init__.py 100.00% <0.00%> (ø)
mmdet/models/backbones/pvt.py 78.15% <0.00%> (ø)
mmdet/models/detectors/lad.py 34.28% <0.00%> (ø)
mmdet/models/backbones/swin.py 81.50% <0.00%> (ø)
mmdet/models/detectors/solo.py 83.33% <0.00%> (ø)
... and 8 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5e246d5...348ce0e. Read the comment docs.

@AronLin AronLin changed the title Support setting pad_val in batch collating [Fix] Support setting pad_val in batch collating Dec 21, 2021
@AronLin AronLin requested a review from hhaAndroid December 22, 2021 09:27
@AronLin
Copy link
Contributor Author

AronLin commented Dec 23, 2021

PanopticFPN: 40.1PQ.
HTC: bbox mAP:42.2, segm mAP:37.4.
They are normal.

@AronLin AronLin requested a review from ZwwWayne December 23, 2021 08:49
@ZwwWayne ZwwWayne merged commit eb33e28 into open-mmlab:dev-v2.20.0 Dec 24, 2021
@AronLin AronLin mentioned this pull request Dec 27, 2021
@AronLin AronLin deleted the pad branch January 18, 2022 10:02
ZwwWayne pushed a commit that referenced this pull request Jul 18, 2022
* Support setting pad_val during collating

* Modify comments
ZwwWayne pushed a commit to ZwwWayne/mmdetection that referenced this pull request Jul 19, 2022
* Support setting pad_val during collating

* Modify comments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants