jin-s13
commented
4 years ago Hi, Thanks for asking! Yes, some annotations may have float visibility. Because in our annotation steps, before manual inspection&correction, we used on-the-shelf models to generate the psudo-labels. The predicted visibility scores are not deleted, which may be useful for further possible applications. In our experiments, we just ignore these keypoints with visibility=0, during training/testing. If the visibility of a keypoint > 0, you can safely use it to train the model.
Hi, thank you very much for making this dataset. I've got a question about annotations.
Some of the annotations (for example, annotation id=193390, lefthand_kpts) have float numbers for the visibility flag instead of integers as described in the DATA_FORMAT file. How to interpret these? Thanks.