No residual connections

[Sooryakiran]

The original paper mentions a residual connection in the generator, and probably I guess that is why you used tanh instead of sigmoid in the generator output. But the given code in the file
SinGAN/models.py
does not seem to contain any residual connections.

[Sooryakiran]

Training with residual connections gives comparable results with just half the number of steps per scale.

[tamarott]

line 63 in SinGAN/models.py: "return x+y" performs summation between the learned output to the clean input image, as described in the paper (see figure 5: https://arxiv.org/pdf/1905.01164.pdf)

[Sooryakiran]

Thank you for the clarification.