pixel-rnn-tensorflow icon indicating copy to clipboard operation
pixel-rnn-tensorflow copied to clipboard

Sequential generation error

Open carpedm20 opened this issue 9 years ago • 3 comments

Current implementation successfully works as a autoencoder but it can't generate from the beginning, from a blank image. This error should be related to masking.

carpedm20 avatar Jul 21 '16 17:07 carpedm20

image

(Up: ground truth, Down: prediction)

Fix with 4c459c11394657c60b12ca0371aadded5e1f208b

carpedm20 avatar Jul 22 '16 12:07 carpedm20

image

Even the one-shot recovering of the original image can be learned quickly as shown in previous comment, the sequential sampling of an image is still not working like igul222/pixel_rnn.

carpedm20 avatar Jul 23 '16 01:07 carpedm20

Update.

generation_2016_08_01_16_40_28

Trained model with python main.py --data=mnist --model=pixel_cnn

carpedm20 avatar Aug 02 '16 01:08 carpedm20