what are the ground-truth images ?

[yayeoCddy]

How we get the ground-truth images, when computing the Quantitative comparison of L2 error ? and Where can we get the supplementary paper of tagan?

[woozzu]

@yayeoCddy When we compute L2 error, we use reconstructed images for positive pairs of image and text. In other words, we forward an image and its corresponding text, then compute L2 error between the input image and a generated image. You can find the supplementary material on our website.