[Feature] Add scale_as_level for multi-level flow loss

[MeowZheng]

Motivation

Optical flow models always predict values of flow as ground truth resolution, even though the resolution is different at each level. This scale_as_level works for deciding whether to scale flow for each level at its native spatial resolution.

Modification

1. Add scale_as_level in multi_level_flow_loss
2. Add scale_as_level in MultiLevelEPE
3. Add scale_as_level in class MultiLevelCharbonnierLoss
4. Revise UT for losses function.

[codecov]

Codecov Report

Merging #58 (a0524e9) into master (55c4625) will increase coverage by 0.14%.
The diff coverage is 80.00%.

@@            Coverage Diff             @@
##           master      #58      +/-   ##
==========================================
+ Coverage   67.09%   67.23%   +0.14%     
==========================================
  Files          83       83              
  Lines        4708     4716       +8     
  Branches      677      677              
==========================================
+ Hits         3159     3171      +12     
+ Misses       1446     1442       -4     
  Partials      103      103              

Flag	Coverage Δ
unittests	67.23% <80.00%> (+0.14%)	⬆️

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

Impacted Files	Coverage Δ
...mflow/models/losses/multilevel_charbonnier_loss.py	91.66% <66.66%> (+0.49%)	⬆️
mmflow/models/losses/multilevel_epe.py	78.04% <66.66%> (+1.12%)	⬆️
mmflow/models/losses/multilevel_flow_loss.py	97.22% <100.00%> (+0.34%)	⬆️
mmflow/datasets/pipelines/transforms.py	79.83% <0.00%> (+0.66%)	⬆️

---

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 55c4625...a0524e9. Read the comment docs.