PixelCNN
PixelCNN copied to clipboard
horizontal and vertical stack
Hi, in the paper, they mentioned that they used a vertical and horizontal convolution stack to overcome blindspot. Was this also included in your implementation?