Raise ImportError on importing malformed COCO directory

[vinnamkim]

Summary

• Previously, COCOImporter implicitly set its rootpath="" and images_dir="" unless the directory structure is not well formed. This makes users cannot know what's happening although their dataset structure is malformed. As a result, after importing, users try to load its images but fail and don't know why.
• This patch makes COCO directory structure rule more strictly at the dataset import level to prevent misbehavior in the next steps.
• In addition, add a checker to check whether Image has correct dimensions: 2 or 3. There have been many test cases which let Image() not having 3 dimensions. This PR also fixes it.

How to test

I added a unit test for it.

Checklist

• I submit my changes into the develop branch
• I have added description of my changes into CHANGELOG
• I have updated the documentation accordingly
• I have added tests to cover my changes
• I have linked related issues

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) 2021 Intel Corporation
#
# SPDX-License-Identifier: MIT

[sooahleex]

It looks good for me. I just have one question, you tested some case for wrong structure like # Wrong structure: ./annotations -> ./labels or # Wrong structure: ./images -> ./imgs, then do we need to consider those kinds of various wrong structure?

[vinnamkim]

It looks good for me. I just have one question, you tested some case for wrong structure like # Wrong structure: ./annotations -> ./labels or # Wrong structure: ./images -> ./imgs, then do we need to consider those kinds of various wrong structure?

It looks good for me. I just have one question, you tested some case for wrong structure like # Wrong structure: ./annotations -> ./labels or # Wrong structure: ./images -> ./imgs, then do we need to consider those kinds of various wrong structure?

In those cases, this PR wants to raise DatasetImportError to tell a user that you should follow COCO directory structure rule. After this PR, the only directory structure exactly fitted to COCO directory structure rule will be accepted.

[codecov-commenter]

Codecov Report

Base: 77.62% // Head: 77.62% // Decreases project coverage by -0.01% ⚠️

Coverage data is based on head (232b2f6) compared to base (f25f2d8).
Patch coverage: 89.47% of modified lines in pull request are covered.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #812      +/-   ##
===========================================
- Coverage    77.62%   77.62%   -0.01%     
===========================================
  Files          184      184              
  Lines        23554    23563       +9     
  Branches      4879     4880       +1     
===========================================
+ Hits         18284    18290       +6     
- Misses        4198     4200       +2     
- Partials      1072     1073       +1     

Flag	Coverage Δ
macos-11_Python-3.8	77.60% <89.47%> (-0.01%)	⬇️
ubuntu-20.04_Python-3.8	77.60% <89.47%> (-0.01%)	⬇️
windows-2019_Python-3.8	77.53% <89.47%> (-0.01%)	⬇️

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

Impacted Files	Coverage Δ
datumaro/plugins/data_formats/coco/base.py	91.30% <85.71%> (-0.67%)	⬇️
datumaro/components/errors.py	87.17% <100.00%> (+0.06%)	⬆️
datumaro/components/media.py	82.40% <100.00%> (+0.15%)	⬆️
datumaro/plugins/ndr.py	75.00% <0.00%> (-1.09%)	⬇️
datumaro/plugins/splitter.py	88.66% <0.00%> (+0.22%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.