[data request] Human 3.6M

[cyfra]

• Name of dataset: Human 3.6M
• URL of dataset: http://vision.imar.ro/human3.6m/description.php
• License of dataset: GRANT OF LICENSE FREE OF CHARGE FOR ACADEMIC USE ONLY
• Short description of dataset and use case(s):

3.6 million 3D human poses and corresponding images

Folks who would also like to see this dataset in tensorflow/datasets, please thumbs-up so the developers can know which requests to prioritize.

And if you'd like to contribute the dataset (thank you!), see our guide to adding a dataset.

[SoumyadeepJana]

@rsepassi I want to contribute to this issue.Consider assigning me for this dataset.

[cyfra]

@KumarArindam - sure, we'll assign it as soon as you accept the collaborator invite.

[SoumyadeepJana]

@cyfra I think KumarArindam had been assigned the humans kinetics dataset.

[KumarArindam]

I would love to collaborate with @SoumyadeepJana on this issue.

[SoumyadeepJana]

@KumarArindam Would love to collaborate and work :)

[KumarArindam]

Sure thing @SoumyadeepJana .Would be happy to. It would be very helpful if you could tell me what work have you done so that we can continue from there.

[jackd]

Sorry I didn't see this earlier. I'm part-way through porting across the data/preprocessing code for this basic baseline paper that looks at simply lifting 2D poses to 3D (so no video/images). The authors provide 2D data from ground truth and 2 different networks. It doesn't handle any video processing so it's a very different problem to the normal dataset.

I've just pushed what I've got to this branch - still needs a bit of work, but I've set up the skeleton.

I'm sure there'll be a some shared components like:

• subject/camera ids
• camera parameters;
• any geometric transformation code we want to provide (frame transforms, projection etc.)
• joint indices/limb connections, etc but if you're looking at doing the video portion I don't think we'll be stepping on each others toes too much.

Feel free to check out what's over there and let me know if there's enough cross-over that you'd like to collaborate.