Release OmniSource ckpts

[kennymckormick]

No description provided.

[codecov]

Codecov Report

Merging #215 into master will decrease coverage by 0.07%.
The diff coverage is 50.00%.

@@            Coverage Diff             @@
##           master     #215      +/-   ##
==========================================
- Coverage   84.39%   84.32%   -0.08%     
==========================================
  Files          78       78              
  Lines        4947     4937      -10     
  Branches      781      782       +1     
==========================================
- Hits         4175     4163      -12     
- Misses        643      644       +1     
- Partials      129      130       +1     

Flag	Coverage Δ
#unittests	84.32% <50.00%> (-0.08%)	⬇️

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

Impacted Files	Coverage Δ
mmaction/datasets/activitynet_dataset.py	95.04% <0.00%> (-1.93%)	⬇️
...maction/models/localizers/utils/post_processing.py	100.00% <ø> (ø)
mmaction/models/localizers/bmn.py	99.00% <100.00%> (+<0.01%)	⬆️
mmaction/models/localizers/bsn.py	100.00% <100.00%> (ø)
mmaction/models/localizers/base.py	54.71% <0.00%> (-2.43%)	⬇️
mmaction/models/recognizers/base.py	60.81% <0.00%> (-2.01%)	⬇️
mmaction/datasets/base.py	67.24% <0.00%> (-1.10%)	⬇️
mmaction/models/losses/base.py	90.00% <0.00%> (-0.91%)	⬇️
mmaction/models/heads/base.py	85.71% <0.00%> (-0.78%)	⬇️
mmaction/models/losses/ohem_hinge_loss.py	100.00% <0.00%> (ø)

---

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 1e77b0b...5be1eee. Read the comment docs.

[kennymckormick]

@innerlee seems OK

[innerlee]

what's the purpose of the projects folder? does it accept public pull requests?
will it be an exhibition?

what do you think? cc. @hellock

[innerlee]

i'm thinking that, how about put the configs and readme under configs/weaklysupervised/omnisource?

although the method might have used slowonly as the backbone, but it does not really belongs to the work of slowfast.
what do you think, @kennymckormick

[kennymckormick]

Actually, in this PR, we only release checkpoints of OmniSource. The training configs are not included (since we haven't released data & code). I think it's OK to put those checkpoints in the TSN and SlowOnly directory, since they are the architectures of these checkpoints. If we put them in another directory: 1. We do not have training configs now. 2. Few people will walk in that directory and see our work.

[innerlee]

okay, then how about put the readme in configs folder (and leave the inference configs where they are now)?

[kennymckormick]

All right, I plan to put the readme in the configs/OmniSource folder, add training configs to that folder after we support OmniSource training in MMAction2