[Enhance] Support 2D&3D Optical Flow Training

[makecent]

Motivation

Fixing the error when decoding optical flows as now the Normalize pipeline is no longer used. See issue #2630

Modification

Modify the RawFrameDecode transform to stack x_frame and y_frame, making the variable imgs of optical flow consistent with RGB, albeit in channel dimension 2 vs 3.

[Dai-Wenxun]

LGTM

[codecov]

Codecov Report

Patch coverage: 62.50% and project coverage change: -0.01% ⚠️

Comparison is base (c548788) 76.18% compared to head (2f692fc) 76.17%.
Report is 5 commits behind head on dev-1.x.

❗ Current head 2f692fc differs from pull request most recent head 1c14912. Consider uploading reports for the commit 1c14912 to get more accurate results

Additional details and impacted files

@@             Coverage Diff             @@
##           dev-1.x    #2631      +/-   ##
===========================================
- Coverage    76.18%   76.17%   -0.01%     
===========================================
  Files          170      170              
  Lines        13792    13780      -12     
  Branches      2361     2360       -1     
===========================================
- Hits         10507    10497      -10     
+ Misses        2718     2716       -2     
  Partials       567      567              

Flag	Coverage Δ
unittests	76.17% <62.50%> (-0.01%)	⬇️

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

Files Changed	Coverage Δ
mmaction/apis/inference.py	79.59% <28.57%> (-3.21%)	⬇️
mmaction/models/losses/hvu_loss.py	86.36% <50.00%> (+0.20%)	⬆️
mmaction/datasets/transforms/formatting.py	92.12% <100.00%> (-0.52%)	⬇️
mmaction/datasets/transforms/loading.py	81.73% <100.00%> (+0.40%)	⬆️
mmaction/models/heads/tpn_head.py	90.69% <100.00%> (+4.65%)	⬆️

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