pose-format
pose-format copied to clipboard
fix masking issue in interpolate
trafficstars
I have found this strange bug: when the x, y, z, and confidence are all 0s for a point in a pose (everything should be masked), but the current implementation on line 342 would not consider it masked when reading the first dimension (the x is 0).
I did not have this issue before, so I assume the pose data should always have masked x, y, and z values when the confidence is 0, @AmitMY is it what we should expect for a "normal" pose?
- I am not sure i understand the change.
- if confidence is 0, a mask is created https://github.com/sign-language-processing/pose/pull/161/files#diff-b2fa5baab13397a32417b3f212c9429c3398ddf14222e50a8eafa2d1af5f379fR50
ready for review and merge