I know the 11 frames' weights' are for the NTIRE 2019 Challenge‘s VSR. What's are your training input for these 11 frames tasks. You use more other dataset except for REDS, or only REDS. Hope that you could show more details about this training process, Very THANKS!!