Some doubts about the experiment mentioned in the paper

[AltoriaDD]

thank you very much for your article C2F-FWN, which has made me gain a lot in human movement transfer. But I still have some doubts in some aspects. Specifically, in the Baselines of Experiments in the paper, you mentioned that each character uses 3000 frames to train EDN, but the video frames of each person wearing the same clothes in the IPER data set do not reach 3000 frames even if they are added up. I want to know how you solve this contradiction in the training and testing stages. I hope you can solve my doubts and tell me more details about this. Thank you very much!