Muhammed Can Keleş

Thank you for your reply. When I run inference, I get 3 sets of anchor points with shapes [1, 1, 3, 3, 1], [1, 5, 1, 3, 1], [1, 12,...

I tried running inference on some H36M frames with the indexing you provided but the results I am getting are still not really meaningfull. I guess I am missing some...

It seems like the multilabel_soft_margin_loss is already implemented, but the task is still in the ToDo list as undone.