Genetic Album Pages:
Evolving Personalized Page Layouts for Visual Content

Joe Geigel
Department of Computer Science
Rochester Institute of Technology


In this talk, I describe a system for generating a set of personalized album pages for a collection of images and other visual content. The system is divided into three modules - a preprocessing module that reorganizes the content by event and sub-event and assigns an "emphasis" value to each piece of content; a page creator module, which is responsible for distributing the content amongst a set of album pages; and a placement module that creates layouts for each individual album page.

At the heart of the last two modules is a genetic engine that utilizes a genetic algorithm (GA) to evolve sets of candidate solutions for each respective problem. Each solution is evaluated based on basic graphic design principles (e.g., balance, emphasis, symmetry, and unity) and the GA is tuned to converge on a set of pages that match predefined user preferences.

The system has been successfully incorporated into a Web-based prototype photo albuming application for interactive page layout over the Internet.

Colloquia Series page.