Boost segmentation import performance

[wonjuleee]

Summary

When analyzing the import performance for cityscapes and kaggle_image_mask, I have checked that the most bottleneck is np.unique for parsing the unique class indices within each mask.
Analysis before PR:

Analysis after PR:

Instead of parsing unique class indices within a mask, I have changed to use all class indices in a dataset.
As a result, the performance is 5 times faster.

How to test

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have added the description of my changes into CHANGELOG.​
• I have updated the documentation accordingly

License

• I submit my code changes under the same MIT License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below).

# Copyright (C) 2023 Intel Corporation
#
# SPDX-License-Identifier: MIT

[jihyeonyi]

Are you going to fix the unit-test errors? It seems that the errors are not directly related to this PR. Your code change looks good to me.

[wonjuleee]

Are you going to fix the unit-test errors? It seems that the errors are not directly related to this PR. Your code change looks good to me.

The unit tests are broken because of checking is_crowd logic in

datumaro/src/datumaro/plugins/data_formats/cityscapes.py

Lines 276 to 283 in 76fc941

	if segm_id < 1000:
	label_id = segm_id
	is_crowd = True
	ann_id = None
	else:
	label_id = segm_id // 1000
	is_crowd = False
	ann_id = segm_id % 1000

.

From my investigation, this is not defined in the official cityscapes document https://www.cityscapes-dataset.com/dataset-overview/#class-definitions, while this is for support CVAT cityscapes format as described in https://opencv.github.io/cvat/docs/manual/advanced/formats/format-cityscapes/.
So, we need to decide whether drop this logic in Datumaro or not.
What do you think @openvinotoolkit/datumaro-maintainers?

[jihyeonyi]

Are you going to fix the unit-test errors? It seems that the errors are not directly related to this PR. Your code change looks good to me.

The unit tests are broken because of checking is_crowd logic in

datumaro/src/datumaro/plugins/data_formats/cityscapes.py

Lines 276 to 283 in 76fc941

	if segm_id < 1000:
	label_id = segm_id
	is_crowd = True
	ann_id = None
	else:
	label_id = segm_id // 1000
	is_crowd = False
	ann_id = segm_id % 1000

.
From my investigation, this is not defined in the official cityscapes document https://www.cityscapes-dataset.com/dataset-overview/#class-definitions, while this is for support CVAT cityscapes format as described in https://opencv.github.io/cvat/docs/manual/advanced/formats/format-cityscapes/. So, we need to decide whether drop this logic in Datumaro or not. What do you think @openvinotoolkit/datumaro-maintainers?

How about marking skip or xfail for this test and investigate this further later?
It is because we could miss something.

[wonjuleee]

Are you going to fix the unit-test errors? It seems that the errors are not directly related to this PR. Your code change looks good to me.

The unit tests are broken because of checking is_crowd logic in

datumaro/src/datumaro/plugins/data_formats/cityscapes.py

Lines 276 to 283 in 76fc941

	if segm_id < 1000:
	label_id = segm_id
	is_crowd = True
	ann_id = None
	else:
	label_id = segm_id // 1000
	is_crowd = False
	ann_id = segm_id % 1000

.
From my investigation, this is not defined in the official cityscapes document https://www.cityscapes-dataset.com/dataset-overview/#class-definitions, while this is for support CVAT cityscapes format as described in https://opencv.github.io/cvat/docs/manual/advanced/formats/format-cityscapes/. So, we need to decide whether drop this logic in Datumaro or not. What do you think @openvinotoolkit/datumaro-maintainers?

How about marking skip or xfail for this test and investigate this further later? It is because we could miss something.

I have fixed unit tests within cityscapes. Please review again :)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (8fdb120) 80.60% compared to head (3317081) 80.60%.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #1261   +/-   ##
========================================
  Coverage    80.60%   80.60%           
========================================
  Files          270      270           
  Lines        30347    30350    +3     
  Branches      5904     5906    +2     
========================================
+ Hits         24462    24465    +3     
  Misses        4504     4504           
  Partials      1381     1381           

Flag	Coverage Δ
ubuntu-20.04_Python-3.8	80.59% <100.00%> (+<0.01%)	⬆️
windows-2022_Python-3.8	80.57% <100.00%> (+<0.01%)	⬆️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jihyeonyi]

LGTM