Towards Book Cover Design via Layout Graphs

Book covers are intentionally designed and provide an introduction to a book. However, they typically require professional skills design produce the cover images. Thus, we propose generative neural network that can book based on easy-to-use layout graph. The graph contains objects such as text, natural scene objects, solid color spaces. This is embedded using convolutional then used with mask proposal generator bounding-box filled object generator. Next, compiled into single image entire trained combination of adversarial training, perceptual reconstruction. Finally, Style Retention Network (SRNet) transfer learned font style onto desired text. Using proposed method allows for easily controlled unique covers.