Retinal image synthesis from multiple-landmarks input with generative adversarial networks