[Enhance] smooth 2D instead of 3D keypoints in body3d_two_stage_video_demo.py

[darcula1993]

smooth 2d results rather than 3d results show better performance

Motivation

smooth 2d results rather than 3d results show better performance

Modification

smooth 2d results before 2d-to-3d lift process.

BC-breaking (Optional)

Use cases (Optional)

Checklist

Before PR:

• I have read and followed the workflow indicated in the CONTRIBUTING.md to create this PR.
• 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:

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

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Merging #1479 (d3d2328) into master (bf9bd53) will decrease coverage by 0.00%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1479      +/-   ##
==========================================
- Coverage   84.47%   84.46%   -0.01%     
==========================================
  Files         236      236              
  Lines       20036    20036              
  Branches     3602     3602              
==========================================
- Hits        16925    16924       -1     
  Misses       2231     2231              
- Partials      880      881       +1     

Flag	Coverage Δ
unittests	84.37% <ø> (ø)

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

Impacted Files	Coverage Δ
mmpose/datasets/pipelines/shared_transform.py	83.26% <0.00%> (-0.43%)	⬇️

---

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 eeebc65...d3d2328. Read the comment docs.

[ly015]

Thank you very much for your help. Could you please sign the CLA?

Do you think it would be better to smooth both 2D and 3D keypoints? Or it's good enough to apply smoothing only on 2D keypoints.

[darcula1993]

Thank you very much for your help. Could you please sign the CLA?

Do you think it would be better to smooth both 2D and 3D keypoints? Or it's good enough to apply smoothing only on 2D keypoints.

According to my test, smooth 2d results with one euro filter config
filter_cfg = dict( type='OneEuroFilter', min_cutoff=0.00001, beta=0.007, )
is good enough.
here is the demo video link:https://mega.nz/folder/3eZVXQqa#6wquV75L-Sdc9Hf9ZmM3Gg
I use faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth for detection and hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth for 2d pose,and videopose3d 1 frame model with retrain weights by myself.