Adrien Bousseau winner of the Young Researcher Award
The ANR (French National Research Agency) has just given its Young Researcher Award to Adrien Bousseau, a researcher in the GraphDeco project-team at Inria Sophia Antipolis – Méditerranée for the DRAO project on digital drawing. This work allows computer vision and 3D image synthesis to be used to automate the creation of drawings while preserving their expressiveness.
DRAO is a research project dedicated to the creation of drawings. The first objective is to gain an understanding of the drawing creation process via user surveys and interviews with professionals. The second objective is the automation of certain steps in the creation of drawings. The last objective is to teach drawing using digital tools. This is a joint project between several Inria research teams, combining Human Machine Interaction with Graphical Computing: GraphDeco (Inria Sophia Antipolis), Manao (Inria Bordeaux), In-Situ (Inria Saclay), Potioc (Inria Bordeaux).
Adrien Bousseau is a digital imaging specialist, who has already been awarded a 2016 Starting Research Position grant by the European Research Council (ERC). Image synthesis lies at the intersection of several scientific fields: computing, mathematics, and physics, whilst examining very concrete problems. Adrien chose this field right out of high school, and he is still pursuing it through the DRAO project.
Could you give us a brief introduction to the DRAO project and how it fits in with your other research work?
Drawing is a powerful creative and communication tool. But producing effective drawings takes time and skill, which slows down the creation process. Drawing is also a hobby enjoyed by many amateurs, but a lot of us don't feel able to produce good drawings. The aim of the DRAO projects was to develop digital drawing tools to facilitate and accelerate the work of professional designers and illustrators, and to help beginners learn how to draw. This project was conducted jointly with HMI (Human Machine Interface) researchers, particularly Theophanis Tsandilas and Wendy Mackay (EX-SITU project-team) from the Inria Saclay Research Center, enabling me to publish my first articles in that field.
What are the special features of the DRAO project, which the ANR singled out as having an "important and original" approach?
The existing digital design tools are based on traditional metaphors where the user draws on a virtual sheet using virtual pencils and colors. Creating quality illustrations with these tools requires the same skills as traditional drawing. This approach is in contrast to the automation offered by 3D image synthesis, which simulates realistic images from virtual scenes, or computer vision, which estimates 3D scenes based on images. The originality of our work lies in the use of computer vision and 3D image synthesis to automate or guide the creation of drawings whilst preserving the advantages of drawing: its expressiveness and flexibility.
We identified three sub-objectives to achieve this. First, we aimed to understand how people draw using the existing tools. To do this, we met professional illustrators and we studied books and tutorials that explain traditional drawing techniques. By understanding those techniques, we were able to suggest new drawing aid algorithms, particularly to estimate 3D information from a drawing and use that information to colorize it (see Figure 1 below).
Our tool automates traditional drawing techniques to quickly colorize a drawing with material effects, such as shiny bodywork, chrome trim, and windshield.
We have also suggested an interface that guides beginners through the basics of drawing by observation (see Figure 2 below).
Our interface helps users draw an object, starting with the general shape before adding details.
You recently presented your work at a specialist HMI conference?
The DRAO project has allowed me to work with researchers who have complementary expertise in image synthesis, computer vision, and human-machine interfaces. I am very happy that we published our results at the best conferences in those three fields, including ACM CHI in the field of human-machine interfaces. Our article describes an observational study that aims for a better understanding of how beginners generate, explore and communicate design ideas using traditional tools. To do this, we have organized two days of design competitions where around ten participants had to perform every stage of a design project, from brainstorming to the manufacture of a prototype, and including the preparatory drawings. Our observations allowed us to suggest objectives to be considered for the development of CAD tools to suit beginner users.