[Feature] Add algebraic triangulation (multi-view 3D pose)

[kristijanbartol]

Motivation

Based on a current roadmap (#9), we want to add "Learnable Triangulation of Human Pose" (link to paper).

Modification

For now, I created a new class in models->detectors->multiview_pose.py.

Use cases (Optional)

The algebraic triangulation is one of the two proposed models (the other is volumetric triangulation). The volumetric triangulation was the state-of-the-art of 3D human pose estimation task for a relatively long time. The algebraic triangulation is simpler, but still powerful model. Moreover, the volumetric triangulation depends on the algebraic triangulation (needs algebraic model to locate the root joint). Therefore, it is reasonable to first add the algebraic triangulation.

Checklist

Before PR:

• [X] I have read and followed the workflow indicated in the CONTRIBUTING.md to create this PR.
• [X] Pre-commit or linting tools indicated in CONTRIBUTING.md are used to fix the potential lint issues.
• [ ] Bug fixes are covered by unit tests, the case that causes the bug should be added in the unit tests.
• [ ] New functionalities are covered by complete unit tests. If not, please add more unit tests to ensure correctness.
• [ ] The documentation has been modified accordingly, including docstring or example tutorials.

After PR:

• [x] CLA has been signed and all committers have signed the CLA in this PR.

[CLAassistant]

All committers have signed the CLA.

[codecov[bot]]

Codecov Report

Merging #1412 (8a93262) into master (c8e91ff) will decrease coverage by 0.00%. The diff coverage is 75.00%.

@@            Coverage Diff             @@
##           master    #1412      +/-   ##
==========================================
- Coverage   84.35%   84.34%   -0.01%     
==========================================
  Files         232      232              
  Lines       19279    19283       +4     
  Branches     3468     3469       +1     
==========================================
+ Hits        16262    16265       +3     
- Misses       2152     2153       +1     
  Partials      865      865              

Flag	Coverage Δ
unittests	84.28% <75.00%> (-0.01%)	:arrow_down:

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

Impacted Files	Coverage Δ
mmpose/models/detectors/multiview_pose.py	69.73% <75.00%> (+0.06%)	:arrow_up:

---

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 c8e91ff...8a93262. Read the comment docs.

[jin-s13]

@kristijanbartol
How is the project going? Is there anything we can help you with?

[kristijanbartol]

Hi @jin-s13 , thank you for your assistance and I apologize for inactivity. I planned to have time for this, but got burried under other obligations. I postponed it for the end of August (ready for review). Is that acceptable? For now, it is seems clear how to integrate the algebraic model.

[jin-s13]

Thank you! That's ok. We look forward to your contributions! Please don't hesitate to contact us if you have any questions.

[ly015]

Hi there, I hope everything is going well. Please let us know if there are any updates or if there is anything we can help with. Thank you very much!

[kristijanbartol]

I'm sorry, guys. I have to postpone this for later. It is still on my roadmap, but I have some higher priority tasks in the next few months.

[ly015]

There is no hurry :) BTW, we have pre-released MMPose v1.0 which involves a major refactoring and is in open beta now. Feel free to have a try if you still work on pose-related projects and any feedback will be appreciated! Thank you