Humans3.6M dataset

[driptaRC]

Can you provide pointers on how to prepare the Humans3.6M dataset?

[wangkaihong]

Hi,

We don't really have a document for that; we just follow what RegDA did according to the information in their repo.

Here is how you should organize the data once you finish downloading:

https://github.com/VisionLearningGroup/UDA_PoseEstimation/blob/master/lib/datasets/human36m.py#L33

[driptaRC]

Assuming I'm downloading from the official site (http://vision.imar.ro/human3.6m/description.php), I get the videos and the poses in .cdf files. Is there a specific way you generate the .json files? Or are you using any script to generate these files?

[wangkaihong]

Sorry we don't remember the detailed downloading process, but we didn't use any script to generate JSON files.