Prepare UI for attributes configuration

[azhiv]

Motivation and context

This PR allows attributes mapping that belong to certain labels (model and task) with ability to pre-map attributes based on their names.
The values returned from nuclio function are also processed during inference to filter out attributes with incompatible types/values.

How has this been tested?

Checklist

• I submit my changes into the develop branch
• I have added a description of my changes into CHANGELOG file
• I have updated the documentation accordingly
• I have added tests to cover my changes
• I have linked related issues (read github docs)
• I have increased versions of npm packages if it is necessary (cvat-canvas,
  cvat-core, cvat-data and cvat-ui)

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) 2022 Intel Corporation
#
# SPDX-License-Identifier: MIT

[azhiv]

@nmanovic @bsekachev Can you please review the approach for automapping label attributes? This is a follow up PR for functionality developed in scope of #3917
@mikhail-treskin

[nmanovic]

@azhiv , is it ready? If so, please remove 'draft' mode. Also could you please look at linter issues?

Indeed I believe the problem with copyright header: Copyright (C) 2019-2021 Intel Corporation. Need to update the year: Copyright (C) 2019-2022 Intel Corporation. If you setup all VS code extensions, it will be updated automatically for files, which you touch. See the list of extensions here: https://openvinotoolkit.github.io/cvat/docs/contributing/development-environment/

[nmanovic]

@klakhov , could you please look at the PR?

[azhiv]

@klakhov I just finished the implementation, can you please trigger the approval checks?

[klakhov]

Hey, great work! That feature looks nice to me.

But while testing this pr I've encountered detection error

I've created a label named label1 with some attributes, and mapped face-detection-0205 label face and it's attributes to my label attrs. Then when i'm trying to annotate image Cant read property 'find' of undefined occurs. It happens on line 1089, where model.attributes[data.label].find is called. I suppose the problem here is that data.label is my label1 and not model's face

Also, could you please update CHANGELOG.md with a link to your pr in Added section and update cvat-ui package version with npm version minor?

[azhiv]

@klakhov A big thanks for your feedback! In fact I missed this case during testing, focusing only on "positive" test cases. I'll fix this issue.
BTW is there a proper place to add tests for this logic?

[klakhov]

@azhiv Currently we don't test 'AI tools' functionality due to CI reasons, so it's better to check as many cases as possible manually

[azhiv]

@klakhov I fixed the issue that you pointed out, and added a small usability enhancement (attributes are automapped during labels selection even if the label names differ). Could you please review the changes?

[klakhov]

Cool, the issue seems to be fixed.

But it seems we have the same problem with attributes. If attribute name is different from model's, even though I mapped them, after annotation is completed my attributes are set to default. I think we should cover this case too.

As you can see on my screenshot, if I'm using label1(with attrs gender, age) everything works fine, and if I'm using label2(with attrs gender123, age123) they are defaulted.

[azhiv]

@klakhov I also noticed this behaviour, but the reason for that is empty attributes array received from the BE

I asked @mikhail-treskin, he needs time to investigate this.

[klakhov]

@azhiv Alright, let us know if you can figure something out 🙂

[azhiv]

@klakhov Please help me to understand the requirements.
Here's the label setup I use for testing:

When I run the automatic annotation against the following configuration:

I get the following result:

The two non-mapped attributes gain default values which is fine, but the first one's value - Adult - is not in the list of allowed ones - [1, 2, 3].

So the question is - what value is attribute1 expected to hold in this scenario? The way it works now doesn't seem correct.

[bsekachev]

@azhiv

So the question is - what value is attribute1 expected to hold in this scenario? The way it works now doesn't seem correct.

I believe if serverless response for an attribute value is out of specification, we should just ignore it. Maybe to show a single warning message in case when we run serverless function on one frame.

[klakhov]

@azhiv , I would agree with @bsekachev, if we have an attribute value that we are not expected to have, it shouldnt be written into mapped value. It would be great to see warning message about such situation just to let user see what happened.

[azhiv]

There's a bunch of lambda manager tests that failed as a result of my recent changes. I was able to localize the problem, but I'd like to get sure and rerun the tests locally before I push the corrected code. How can lambda manager tests be executed in my local environment? This one test_api_v2_lambda_requests_create_empty_mapping for example.

[azhiv]

@bsekachev @klakhov Thanks for your feedback!
Should the gained attribute be rejected as a result of check_attr_value?

[merchantbreeze]

It's a nice work I need!
It's vaild on the jobs page with AI Tools, but when I used the Automatic annotation with your example model, I've encountered detection error.


Here is the logs in cvat

2022-05-19 14:32:30,258 DEBG 'rqworker_low' stderr output:
DEBUG - 2022-05-19 14:32:30,258 - worker - Handling failed execution of job 953d039f-1148-44e3-922b-07e8d5e7a576

2022-05-19 14:32:30,260 DEBG 'rqworker_low' stderr output:
ERROR - 2022-05-19 14:32:30,260 - worker - Traceback (most recent call last):
  File "/home/django/cvat/apps/lambda_manager/views.py", line 197, in invoke
    mapped_attributes = attr_mapping[func_label].get("attributes")
KeyError: 'face'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/opt/venv/lib/python3.8/site-packages/rq/worker.py", line 936, in perform_job
    rv = job.perform()
  File "/opt/venv/lib/python3.8/site-packages/rq/job.py", line 684, in perform
    return self.func(*self.args, **self.kwargs)
  File "/home/django/cvat/apps/lambda_manager/views.py", line 618, in __call__
    LambdaJob._call_detector(function, db_task, labels, quality,
  File "/home/django/cvat/apps/lambda_manager/views.py", line 468, in _call_detector
    annotations = function.invoke(db_task, data={
  File "/home/django/cvat/apps/lambda_manager/views.py", line 241, in invoke
    raise ValidationError(
django.core.exceptions.ValidationError: ["`openvino-omz-face-detection-0205` lambda function was called without mandatory argument: 'face'"]
Traceback (most recent call last):
  File "/home/django/cvat/apps/lambda_manager/views.py", line 197, in invoke
    mapped_attributes = attr_mapping[func_label].get("attributes")
KeyError: 'face'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/opt/venv/lib/python3.8/site-packages/rq/worker.py", line 936, in perform_job
    rv = job.perform()
  File "/opt/venv/lib/python3.8/site-packages/rq/job.py", line 684, in perform
    self._result = self._execute()
  File "/opt/venv/lib/python3.8/site-packages/rq/job.py", line 690, in _execute
    return self.func(*self.args, **self.kwargs)
  File "/home/django/cvat/apps/lambda_manager/views.py", line 618, in __call__
    LambdaJob._call_detector(function, db_task, labels, quality,
  File "/home/django/cvat/apps/lambda_manager/views.py", line 468, in _call_detector
    annotations = function.invoke(db_task, data={
  File "/home/django/cvat/apps/lambda_manager/views.py", line 241, in invoke
    raise ValidationError(
django.core.exceptions.ValidationError: ["`openvino-omz-face-detection-0205` lambda function was called without mandatory argument: 'face'"]

2022-05-19 14:32:30,260 DEBG 'rqworker_low' stderr output:
DEBUG - 2022-05-19 14:32:30,260 - worker - Invoking exception handler <function rq_handler at 0x7f446a60bee0>

2022-05-19 14:32:30,273 DEBG 'rqworker_low' stderr output:
DEBUG - 2022-05-19 14:32:30,273 - worker - Sent heartbeat to prevent worker timeout. Next one should arrive within 480 seconds.

2022-05-19 14:32:30,273 DEBG 'rqworker_low' stderr output:
DEBUG - 2022-05-19 14:32:30,273 - worker - Sent heartbeat to prevent worker timeout. Next one should arrive within 480 seconds.

2022-05-19 14:32:30,274 DEBG 'rqworker_low' stderr output:
DEBUG - 2022-05-19 14:32:30,274 - worker - *** Listening on low...

2022-05-19 14:32:30,275 DEBG 'rqworker_low' stderr output:
DEBUG - 2022-05-19 14:32:30,275 - worker - Sent heartbeat to prevent worker timeout. Next one should arrive within 480 seconds.

It seems like an error about attribute dict, key 'face' is not exist in the dict so we got a KeyError.

[merchantbreeze]

I have solved it.I found that the respond method of Automatic Annotation is different with AI Tool in jobs page, the attrsMapping is missing from the args, so we can't get it.

[bsekachev]

Hi @azhiv

Thank you for the contribution.
We've finished the patch and merged it to our current CVAT fork: https://github.com/cvat-ai/cvat

You can see corresponding commit here :)

[azhiv]

@bsekachev Great news, at last! Thanks for sharing.