Speedup loading large yolo datasets.

[satheeshkatipomu]

Currently, loading yolo dataset(~150k) is taking prohibitively long time compared to loading the same dataset in coco format.

I think it is mostly because of yolo annotation format where we don't have actual image width and heights. So we are trying to get the height and width of the images using imagesize.get but required while converting to other formats.

Nothing much, just parallelized for loop using joblib.Parallel using half the available threads by default. Also updated copyright year in the docs.

Speed test on ~150k dataset:
before joblib.Parallel : >180 mins
after joblib.Parallel: ~12 mins

[codecov-commenter]

Codecov Report

Merging #26 (d105b39) into main (1ed11c4) will increase coverage by 0.29%.
The diff coverage is 95.34%.

@@            Coverage Diff             @@
##             main      #26      +/-   ##
==========================================
+ Coverage   74.98%   75.28%   +0.29%     
==========================================
  Files          18       18              
  Lines        1659     1679      +20     
==========================================
+ Hits         1244     1264      +20     
  Misses        415      415              

Impacted Files	Coverage Δ
optical/converter/yolo.py	93.33% <95.23%> (+1.78%)	⬆️
optical/converter/utils.py	88.98% <100.00%> (+0.04%)	⬆️

---

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 1ed11c4...d105b39. Read the comment docs.

[satheeshkatipomu]

docs build is failing because of jinja2 issue, it goes away after updating Sphinx version to 4.5.0. I have committed these changes, docs build works fine locally but during docs build workflow it is still installing old version(maybe using cache).

Update: Fixed it by updating the versions in docs/requirements.txt apart from pyproject.toml.

[bharatkumarreddy]

Thanks for the PR. Looks good to me

[bishwarup307]

thanks for the PR.