Using fast optimization techniques to fine tune label placement in an automatically generated image

Randomly created question and solution problem sets will frequently include scaled images that are generated with a series of variables. These figures contain labels for such things as dimensions, coordinates, angles, sub-titles and vertex names. When images are created manually, labels are placed in a manner which avoids them from obscuring other labels or the images itself. In an automated setting however, this is not so easily managed. The purpose of this project is to research, design and apply machine learning methodologies for optimal placement of all image labels on a computer-generated image so all labels are near their correct location, clearly identifiable and do not overlap the image or other labels nearby.

Faculty Supervisor:

Osmar Zaiane

Student:

Partner:

Varafy

Discipline:

Computer science

Sector:

Information and cultural industries

University:

University of Alberta

Program: