VIDEO DESCRIPTION:

Play, download and edit the free video [VIS19 Preview] Data Brushes: Interactive Style Transfer for Data Art (visap paper).

VIS Arts Program

Authors: Mahika Dubey, Jasmine Tan Otto, Angus G. Forbes

Abstract: This paper introduces Data Brushes, an interactive web application to explore neural style transfer using models trained on data visualizations. Our application includes two distinct modes that invite casual creators to interact with deep convolutional neural networks to co-create custom artworks. The first mode, ‘magic markers’, mimics painting with a brush on a canvas, enabling users to paint a style onto selected areas of an image. The second mode, ‘compositing stamps’, uses a real-time method for applying style filters to selected portions of an image. Specifically, we focus on style transfer networks created from canonical and contemporary works of data visualization and data art. In addition to enabling a novel cre- ative workflow, the process of interactively modifying an image via multiple style transfer networks reveals meaningful features encoded within the networks, and provides insight into the effect particular networks have on different images, or even different regions of an image, such as border artifacts and spatial stability or instability. To evaluate Data Brushes, we gathered expert feedback from participants of a data science symposium and ran an observational study, finding that our application facilitates the creative exploration of neural style transfer for data art and enhances user intuition regarding the expressive range of particular style transfer features.

Download, play and edit free videos and free audios from [VIS19 Preview] Data Brushes: Interactive Style Transfer for Data Art (visap paper) using RedcoolMedia.net web apps