Can i train this model on other datasets,like ImageNet,voc etc。For the images in these datasets,Do I need to resize the images(for x8、x4)。