From a computer science standpoint, this is an extremely interesting problem. Because football is highly structured, the football domain presents a rich set of challenging AI and Computer Vision problems, the implications of which extend well beyond the scope of American football. For this reason, research stemming from the football domain will be of general interest to the AI and Computer Vision communities.
In addition, the football domain is of significant practical and commercial interest because all professional football teams and nearly all major college teams employ crews of video scouts whose only job is to organize, annotate, and analyze huge collections of football video to provide coaches with information upon which they can base their strategy decisions for both individual games and entire seasons. The software currently used by most of these teams costs tens of thousands of dollars, despite the fact that it is little more than a user interface for annotation and archival and does not do anything “smart.”
The goal of our research is to develop automatic methods to perform many of the tasks that are currently performed manually by video scouts. In the end, we would like to create a full system that can analyze a collection of football video and make strategy recommendations to coaches based on that analysis. However, our work is only just beginning. You can keep track of our progress through the System Overview and Publications pages.