Hi! I already posted about this on the original tensorflow implementation, but i would also like to know your opinion on the matter :) Do you think that the black borders on an image like the one bellow would affect the training and the predicitons of depth/pose? 