rawalkhirodkar/egohumans

Benchmark for evaluating human mesh recovery

Closed this issue · 2 comments

fabienbaradel commented

Hi @rawalkhirodkar,
Thanks for releasing this dataset, this is very interesting and challenging for the research community.
In your paper you are not really mentioning the task of 3d human mesh recovery, do you plan (or do you already have something ready) for benchmarking human mesh recovery methods?
Should we use the same training/testing split as for the 2D pose benchmark? What do you think?
Thanks for you help and your time,
Best

rawalkhirodkar commented

Thank you for your interest.
We have no immediate plans for mesh recovery benchmarking - I would be happy to join forces if you are interested.
For consistency, we should use the same train/test splits.

  • 👍1
fabienbaradel commented

Great to see your interest. I will take a look at your train/test splits from the 2D pose task and will get back to you asap.