Sketchai: using FCNS to Extract Line ART Drawings

Abstract

The digital revolution has improved every field of human lives. And field of ART is no exception. The rapid development of modern technology and techniques has made an impact on the work of painters, sketch artists and even comic book writers. One would be hard pressed to find artists in this day and age who haven’t heard of modern tools and applications such as Adobe Photoshop, GIMP, etc. But though these applications have undoubtedly made the lives of artists better, the use of AI for art is still at a nascent stage. The current work aims at developing a web-based application called SketchAI which uses Artificial Intelligence to simplify rough sketches or art work and extract the simplified line drawing. Fully Convolution Networks or FCNs are used to automate the task of sketch simplification. Dataset of rough sketches and corresponding line art drawings are used to train the system to extract line art. Different parameters such as number of epochs, loss functions etc. are considered for experimentation along with different subsets and augmentations of the data. Finally, comparisons of different methods are done to integrate deep learning models with a web application.