edge2art: Edges to Artworks Translation with Conditional Generative Adversarial Networks

This paper presents an application of the pix2pix model [3], which presents a solution to the image to image translation problem by using cGANs. The main objective of our research consists in the evaluation of several artificial artworks that were generated by the cGAN, taking a scribble of edges as input. This evaluation covers different artistic movements and art styles such as Rococo, Ukiyo-e, Fauvism and Cubism. The set of the trained models of these different styles is called edge2art. Each art style was trained over more than 2000 artworks examples taken from the wikiart dataset used in ArtGAN [4]-[5]. The experiments consists in giving scribbles images to the cGAN, and depending on the selected style, the network will give an colored and stylized artwork as output. Comparison between the generated artworks and the target artworks are measured by Mean Squared Error and Structural Similarity Measure.