CONTOUR GENERATOR AND STYLE TRANSFER OF MURAL PAINTINGS USING CYCLE GAN CONTOUR GENERATOR AND STYLE TRANSFER OF MURAL PAINTINGS USING CYCLE GAN

India is known for its cultural diversity and is home to several heritage sites. There was a period when many art
forms flourished and mural paintings, sculpting, singing and dancing are some of them. Currently, India has more
than 20 locations containing mural painting and most of these are caves and rock cut chambers. These art forms are
local and only a few artisans are skilled enough to replicate or imitate them. Thus a technique to artistically transform
the style of an image is proposed which involves the creation of two models – Contour Generator and Style Transfer
Generator. The paintings are first translated into their respective outlines or sketches. Thus an efficient structural
outline, the Contour, of the mural art can be generated using a Cycle Generative Adversarial Network (Cycle GAN).
Cycle GAN can be used in an image-to-image translation technique for conversion of images between two domains,
here from paintings to sketches. The Style Transfer generator can now transfer any contour into a Mural painting of
this style. This way one can create a mural art with just a sketch of the painting. Also one can transform the style of
a random picture to a Mural art by using the Contour Generator and style transfer generator. The model when
executed successfully will serve for multiple purposes. It can be used to extract the basic structure of an image, fuse
the painting with the other and many more.