Preprocessing of how2sign and phoenix

Hi! Thank you for providing the data. I wanted to ask if the phoenix and how2sign datasets were preprocessed in a similar manner. I noticed that how2sign keypoints are smaller than the ones in phoenix. Were they divided by some value before being saved? Also, did you normalize the keypoints in each dataset (using the shoulder distance etc)? 