Question about goal image selection

[tianhaowuhz]

When using the video prompt for inferencing, as shown in the code, the algo will find the nearest index according to ee pose to determine the next goal. However, if I understand correctly, the nearest index is computed from "goal_ee_traj", which seems to be the "robot ee pos". So if I only have human goal images, how do I determine the next goal image?