We generate uncurated images of the same person using randomly sampled poses.

We compare the results of our method with the pretrained method and the full fine-tuning method.