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.

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

MMOCR COlab #147

Closed
wants to merge 28 commits into from
Closed

MMOCR COlab #147

wants to merge 28 commits into from

Conversation

quincylin1
Copy link
Contributor

Colab for mmocr, including:

Training, testing and visualization for text recognition demo dataset
Testing and visualization for text detection with PANet
Testing and visualization for KIE

@quincylin1 quincylin1 mentioned this pull request Apr 30, 2021
22 tasks
@codecov
Copy link

codecov bot commented Apr 30, 2021

Codecov Report

Merging #147 (f359687) into main (0e70f99) will increase coverage by 1.98%.
The diff coverage is 86.36%.

Impacted file tree graph

@@            Coverage Diff             @@
##             main     #147      +/-   ##
==========================================
+ Coverage   80.23%   82.22%   +1.98%     
==========================================
  Files         107      111       +4     
  Lines        6518     7194     +676     
  Branches      948     1060     +112     
==========================================
+ Hits         5230     5915     +685     
+ Misses       1061     1023      -38     
- Partials      227      256      +29     
Flag Coverage Δ
unittests 82.22% <86.36%> (+1.98%) ⬆️

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

Impacted Files Coverage Δ
mmocr/datasets/pipelines/kie_transforms.py 34.78% <0.00%> (-1.59%) ⬇️
...s/textrecog/recognizer/encode_decode_recognizer.py 92.77% <ø> (ø)
mmocr/models/textdet/dense_heads/fce_head.py 25.00% <25.00%> (ø)
...ls/textdet/detectors/single_stage_text_detector.py 88.88% <50.00%> (-11.12%) ⬇️
mmocr/models/textdet/detectors/fcenet.py 53.84% <53.84%> (ø)
mmocr/models/textrecog/recognizer/base.py 54.11% <57.14%> (-0.32%) ⬇️
mmocr/models/textrecog/losses/ctc_loss.py 78.37% <70.00%> (-4.39%) ⬇️
mmocr/datasets/pipelines/transforms.py 81.54% <86.50%> (+1.81%) ⬆️
...tasets/pipelines/textdet_targets/fcenet_targets.py 93.51% <93.51%> (ø)
mmocr/apis/inference.py 93.65% <95.34%> (+4.17%) ⬆️
... and 14 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 0e70f99...f359687. Read the comment docs.

@innerlee
Copy link
Contributor

Linting fails

@innerlee innerlee mentioned this pull request May 15, 2021
innerlee and others added 25 commits May 17, 2021 11:49
* Improve KIE
* allow data do not contain label
* Make relation float32

Signed-off-by: lizz <lizz@sensetime.com>

* Add test

Signed-off-by: lizz <lizz@sensetime.com>
* Implement iter protocal for Loader

Signed-off-by: lizz <lizz@sensetime.com>

* Test iter protocal

Signed-off-by: lizz <lizz@sensetime.com>
* add model for chinese

* update readme

* update readme

* fix link error; add vis for Chinese
* Add unit test for dbnet transforms

* Add unit test for dbnet transforms

* Add unit test for dbnet transforms
* change argument names according to convention

* bug fix when rename leakyRelu
* fix bug of loss nan

* update

* update ctc loss
Signed-off-by: lizz <lizz@sensetime.com>
* add mmocr demo documentation

* Update text_det_demo.md

* Update text_recog_demo.md

* Update README.md

* Update README.md

* Update README.md

Co-authored-by: Hongbin Sun <hongbin306@gmail.com>
* update demo docs

* update docs
* move demo.md to demo/docs

* update link in md
* Add support for numpy arrays in model_inference

* Add test for numpy ndarray inference

* Fix linting problems

* Add support for batch inference

* Add batch inference demo script

* Fix comment

* Test batch inference with paths and arrays

* lint code

* Update model_inference docstring

* Refactor model inference tests

* Change inference function to make text detectors and recognizers use the same input data types

* Change single state text detector model to support batch inference

* Lint code

* simplify inference tests

* Remove psenet from batch inference test cases to prevent the pytest being killed

* Update batch_image_demo.py

* fix bug when test with dataset 

fix bug when test with dataset, for example, `./tools/dist_test.sh configs/textrecog/sar/sar_r31_parallel_decoder_academic.py <checkpoint> 1 --eval acc`

Co-authored-by: Hongbin Sun <hongbin306@gmail.com>
* refactor tps config

* recover tps

* add ckpt of tps
Signed-off-by: lizz <lizz@sensetime.com>
* Add things from "import *" to __all__

Signed-off-by: lizz <lizz@sensetime.com>

* Remove tile_image

Signed-off-by: lizz <lizz@sensetime.com>
* move demo image to demo/resources

* update index.rst

* revert

* update github link for demo image
* add metafiles

* add metafiles

* add metafiles

* fix metafile errors

* fix metafiles

* Update metafile.yml

Co-authored-by: Hongbin Sun <hongbin306@gmail.com>
* add mminstall.txt

* fix mminstall.txt

* fix mminstall
* fix open-mmlab#173: support aug test

* fix pytest

* support batch inference for both text det and recog

* update unittest

* use one img for batch infer test
* add fcenet

* fix linting and code style

* fcenet finetune

* Update transforms.py

* Update fcenet_r50dcnv2_fpn_1500e_ctw1500.py

* Update fcenet_targets.py

* Update fce_loss.py

* fix

* add readme

* fix config

* Update fcenet_r50dcnv2_fpn_1500e_ctw1500.py

* fix

* fix readme

* fix readme

* Update test_loss.py

Co-authored-by: Hongbin Sun <hongbin306@gmail.com>
jeffreykuang and others added 2 commits May 17, 2021 11:49
* postproc

* mv postproc of textdet in mmcv

* linting

* change deps mmcv 1.3.4

* update deps

* linting

* Update configs/textdet/panet/panet_r18_fpem_ffm_600e_icdar2015.py

* Update docs/changelog.md

* Update mmocr/models/textdet/postprocess/__init__.py

* Update mmocr/version.py

* resize&recompress jpg

* Update changelog.md

Co-authored-by: Hongbin Sun <hongbin306@gmail.com>
@innerlee innerlee closed this May 17, 2021
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.

7 participants