← AttnGAN: Fine-Grained Text to Image Generation …