Revisited: "Video Analytics: From Keywords to Keyframes”

17 Nov, 2016

Revisited: "Video Analytics: From Keywords to Keyframes”

Miyoko Conley, a BCNM Designated Emphasis candidate in Theater, Dance, and Performance Studies, recaps Virginia Kuhn's History and Theory of New Media Lecture Series presentation on "Video Analytics From Keywords to Keyframes", which took place on November 3, 2016.

November 3rd marked the last History and Theory of New Media lecture for the fall 2015 semester, with a stimulating lecture from Dr. Virginia Kuhn from University of Southern California. Entitled “Video Analytics From Keywords to Keyframes,” the lecture covered Dr. Kuhn’s platform Video Analysis Tableau (VAT) and the ethos behind its inception. The VAT applies computational methods to the study of vast video archives.

Before explaining the tool itself, Dr. Kuhn emphasized the larger theoretical concerns driving the project. If we recognize that digital processes are just as important as words and images, and that contemporary literacy does not only include textual literacy, but also digital literacy (and other forms of cultural production), then it is imperative that critical theorists weigh in on these large matters of public concern. It is also important for theorists to recognize that in this digital age, scholarship can take many forms besides the traditional essay format, as Dr. Kuhn’s digital dissertation (one of the first in the United States) and her classes at USC on creating video arguments show.

Dr. Kuhn’s focus on video sheds light on another important aspect of studying new media: the prevalence of moving images in particular. With more footage than one could watch in a lifetime uploaded to YouTube in an hour, and projections that by 2020, 82 percent of all consumer internet traffic will comprise of video, Dr. Kuhn’s argument that without having specific tools to study this media, it will remain invisible to us, easily follows. However, the question of how to study vast amounts of video over a sustained period of time continues to be a problem. Not only are moving images complex on a theoretical level in that they intertwine subjectivity and objectivity, but also there are practical concerns of how much time it takes to break down one video, let alone a whole archive. Added to this mix are the ongoing issues of how make video available for study and educational purposes when current copyright law is still struggling to catch up to the proliferation of material uploaded to privately-owned sites like YouTube. One solution Dr. Kuhn has developed is the VAT, which tackles the question of “how to study?” head-on.

While not all the code has been streamlined as of this writing, the intention of the VAT is to provide educators with a tool for large-scale video analysis that takes out some of the labor usually involved with this type of work. More than that, however, the VAT seeks to combine what Dr. Kuhn calls “human vision” with “computer vision,” meaning, to combine the strength of human vision to focus on a detail, and the strength of computer vision to draw attention to aspects unnoticed. At the moment, this mainly manifests in the different ways the VAT can visualize the digital data, in order to give educators a multilayered approach to studying motion pictures that extends beyond looking frame-by- frame. Please be sure to check out our 2016 spring lectures with Kavita Philip and Eden Medina!