the 43rd international conference and exhibition on
24-28 July
Anaheim, California
The Sketchy Database is the first large-scale collection of sketch-photo pairs. Crowd workers sketch 12,500 objects from 125 categories and generate 74,425 sketches. The system trains cross-domain CNNs to embed sketches and photographs, which leads to state-of-the-art sketch-based image retrieval.
Patsorn Sangkloy
Georgia Institute of Technology
Nate Burnell
Brown University
Cusuh Ham
Georgia Institute of Technology
James Hays
Georgia Institute of Technology
The first line drawing vectorization algorithm that explicitly balances fidelity to the input bitmap with simplicity of the output curves. The global optimization generates few yet accurate curves and disambiguates curve topology at junctions by favoring the simplest interpretations overall.
Jean-Dominique Favreau
INRIA Sophia Antipolis
Florent Lafarge
INRIA Sophia Antipolis
Adrien Bousseau
INRIA Sophia Antipolis
A novel technique for fully automatic simplification of sketch drawings. By learning a series of convolution operators from a dataset of rough and simplified-sketch drawing pairs, this method overcomes the limitation of existing methods that require vector inputs. It outperforms existing approaches in a user study.
Edgar Simo-Serra
Waseda University
Satoshi Iizuka
Waseda University
Kazuma Sasaki
Waseda University
Hiroshi Ishikawa
Waseda University
Introducing a fully automatic method for generation of legible compact calligrams that provides a balance among conveying the input shape, legibility, and aesthetics.
Changqing Zou
Hengyang Normal University, Simon Fraser University
Junjie Cao
Dalian University of Technology, Simon Fraser University
Warunika Ranaweera
Simon Fraser University
Ibraheem Alhashim
Simon Fraser University
Ping Tan
Simon Fraser University
Alla Sheffer
The University of British Columbia
Hao Zhang
Simon Fraser University
This texture-synthesis method for generating believable handwriting replicates the style of a specific author. In user studies, it is capable of fooling casual observers, and it works on historical figures.
Tom Haines
University College London
Oisin Mac Aodha
University College London
Gabriel Brostow
University College London