Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Feature] Support Real-time model ERFNet #960

Merged
merged 16 commits into from
Dec 2, 2021
Merged

[Feature] Support Real-time model ERFNet #960

merged 16 commits into from
Dec 2, 2021

Conversation

MengzhangLI
Copy link
Contributor

@MengzhangLI MengzhangLI commented Oct 13, 2021 •
edited
Loading

ERFNet: Efficient Residual Factorized ConvNet for Real-time Semantic Segmentation

Paper: http://www.robesafe.uah.es/personal/eduardo.romera/pdfs/Romera17tits.pdf (Current Citation: 596)
Repo: https://github.com/Eromera/erfnet_pytorch (Current Star: 346)

Validation Result 72.20 %

These two figures below may be helpful for code review.
Structure of the model:
image

Two main block of the structure (Downsampler Block and non-bt-1D):
image

  • Align Inference metric
  • Add unittest
  • Align Training metric
  • Upload model & log and README.md

@MengzhangLI
Copy link
Contributor Author

MengzhangLI commented Oct 13, 2021 •
edited
Loading

Inference metric has done.

Input torch.ones(1, 3, 512, 1024).cuda()

Original repo is output.max(): 10.0800:
image

My MMSeg repo is also output.max(): 10.0800:
image

@MengzhangLI MengzhangLI added the WIP Work in process label Oct 13, 2021
@codecov
Copy link

codecov bot commented Oct 13, 2021 •
edited
Loading

Codecov Report

Merging #960 (efee39b) into master (91cbe06) will increase coverage by 0.16%.
The diff coverage is 100.00%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #960      +/-   ##
==========================================
+ Coverage   89.57%   89.74%   +0.16%     
==========================================
  Files         120      121       +1     
  Lines        6717     6828     +111     
  Branches     1122     1139      +17     
==========================================
+ Hits         6017     6128     +111     
  Misses        496      496              
  Partials      204      204
Flag Coverage Δ
unittests 89.74% <100.00%> (+0.16%) ⬆️

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

Impacted Files Coverage Δ
mmseg/models/backbones/__init__.py 100.00% <100.00%> (ø)
mmseg/models/backbones/erfnet.py 100.00% <100.00%> (ø)

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 91cbe06...efee39b. Read the comment docs.

@MengzhangLI MengzhangLI self-assigned this Oct 19, 2021
@Junjun2016
Copy link
Collaborator

Hi @MengzhangLI
The CI has been fixed, please merge the master branch.

@MengzhangLI
Copy link
Contributor Author

Refactoring the code has been done.

image

@MengzhangLI
Copy link
Contributor Author

The training metric has been done.
From paper, the mIoU of scratch/ pretrained model on Cityscapes val is 70.0 and 71.5, respectively.

image

Result of scratch ERFNet with FCN head using MMSegmentation is 70.8.

image

@Junjun2016
Copy link
Collaborator

Please fix the conflict.

Junjun2016
Junjun2016 reviewed Nov 8, 2021
View reviewed changes
configs/erfnet/README.md Outdated Show resolved Hide resolved
configs/erfnet/fcn_erfnet_4x4_512x1024_160k_cityscapes.py Outdated Show resolved Hide resolved
configs/erfnet/fcn_erfnet_4x4_512x1024_160k_cityscapes.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
mmseg/models/backbones/erfnet.py Outdated Show resolved Hide resolved
@Junjun2016
Copy link
Collaborator

Junjun2016 commented Nov 9, 2021 via email

@Junjun2016
Copy link
Collaborator

Junjun2016 commented Nov 9, 2021 via email

@Junjun2016
Copy link
Collaborator

Hi @MengzhangLI
Please resolve the conflicts.

@RockeyCoss
Copy link
Contributor

Please merge the master branch into your branch, thank you.

MengzhangLI
MengzhangLI commented Nov 26, 2021
View reviewed changes
configs/erfnet/README.md Outdated Show resolved Hide resolved
Junjun2016
Junjun2016 reviewed Nov 30, 2021
View reviewed changes
README.md Outdated Show resolved Hide resolved
Junjun2016
Junjun2016 reviewed Nov 30, 2021
View reviewed changes
README_zh-CN.md Outdated Show resolved Hide resolved
@xvjiarui
Copy link
Collaborator

xvjiarui commented Dec 1, 2021

Please fix the lint

@xvjiarui xvjiarui merged commit ff1e2d8 into open-mmlab:master Dec 2, 2021
@MengzhangLI MengzhangLI removed the WIP Work in process label Dec 2, 2021
@MengzhangLI MengzhangLI deleted the ERFNet branch February 1, 2022 03:18
bowenroom pushed a commit to bowenroom/mmsegmentation that referenced this pull request Feb 25, 2022
@MengzhangLI
[Feature] Support Real-time model ERFNet (open-mmlab#960)
6a2cfea 
* first commit

* Fixing Unittest Error

* first refactory of ERFNet

* Refactorying NonBottleneck1d Module

* uploading models&logs

* uploading models&logs

* fix partial bugs & typos

* ERFNet

* add ERFNet with FCNHead

* fix typos of ERFNet

* add name on README.md cover

* chane name to T-ITS'2017

* fix lint error
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Junjun2016 Junjun2016 Junjun2016 approved these changes

@xvjiarui xvjiarui xvjiarui approved these changes

Assignees

@MengzhangLI MengzhangLI

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@MengzhangLI @Junjun2016 @RockeyCoss @xvjiarui