Fig. 4
From: Stacked generative adversarial networks for image compositing

Training convergence curves of L2 loss. a FiveK. b AMOS
From: Stacked generative adversarial networks for image compositing
Training convergence curves of L2 loss. a FiveK. b AMOS