pytorch-0.4-yolov3
pytorch-0.4-yolov3 copied to clipboard
training with single image
Would it be possible to have the network to fit to a single image? So, for example, if I just have a single image with a single class instance, could I train the network to just leern this single image class i.e. overfit to this single instance?
Maybe.. But you should apply the excessive data augmentation for single image.