Not sure how you handle multi person

[valentin-fngr]

Hi,

I just wanted to ask how you handle multi person as I cannot figure it out. I do understand that you have (N, nb_joint, H, W) output size. How do you associate a keypoint to an individual ?

Thank you very much

[MiraPurkrabek]

It also needs to be clarified to me from the paper. @bmartacho , could you please elaborate? When I run the inference demo, it estimates only one person in each image. Does that mean the network is a single-person pose estimation network and, therefore, top-down?