Unconditional generation was used, the input to the model was clear images, and after the training was completed the training set plus random noise was used for testing, and it appeared that the test results could not be matched with the labels one-to-one.
Unconditional generation was used, the input to the model was clear images, and after the training was completed the training set plus random noise was used for testing, and it appeared that the test results could not be matched with the labels one-to-one.