Hi @bearpaw
one of my question is about the visibility of joints you created by the joints coordinate , if the coordinate is zeros not-visible , why do not use the is_visible attribute of the dataset itself, Do you want to train on the joints that are not visible but do exist in image, ?
How about 'Stacked hourglass paper' I saw that they used the visibility from dataset but did not find in it's code for training that they consider the occluded joints?
Hi @bearpaw one of my question is about the visibility of joints you created by the joints coordinate , if the coordinate is zeros
not-visible, why do not use theis_visibleattribute of the dataset itself, Do you want to train on the joints that are not visible but do exist in image, ? How about 'Stacked hourglass paper' I saw that they used thevisibility from datasetbut did not find in it's code for training that they consider the occluded joints?I would appreciate your comment.