Large-Scale Video Analysis Project

Large Scale Video Analysis Project

Historically, the analysis of large video archives has been made difficult an inability to sort and categorize video content on a massive scale, further complicated by a lack of metadata and the difficulty of articulating the complexity of sound as text. In response to the inherent problems in the cultural and historical study of large video archives, the Large Scale Video Analysis project (LSVA) seeks to create a framework to study, retrieve, and annotate massive collections of moving image data. Virginia Kuhn of the University of Southern California, along with co-investigators Michael Simeone of the Nexus Lab, Alan Craig of I-CHASS, as well as David Bock, Liana Diesendruck, and Luigi Marini of NCSA, are finding ways of using High Performance Computing (HPC) to compare cross-culture videos for the study of the arts, history, and culture. Their work aims to allow researchers to view “slices” of videos and decipher such unique characteristics as camera angles, movements, and color timing. In application, LSVA can find examples of a camera angle made famous by a well-known director, or a cultural preference for a technique used to record a news interview, for example. The compute-intensive and memory-intensive analysis of LSVA requires HPC platforms, in this instance, the Gordon supercomputer at the San Diego SupercomputingCenter (SDSC). Researchers are currently attempting to find solutions for searching, tagging, and analyzing user-identified objects. They are pursuing a two pronged approach to place more interpretive power in the hands of users and implement a customized on-demand configuration of Gordon that enables iterative queries over a short period of time.

Affiliated Organizations

Project Director(s)

Michael Simeone

Director of Data Science and Analytics, Hayden Library

Headliner Info

Alan Craig, I-CHASS
David Bock, NCSA
Liana Diesendruck, NCSA
Luigi Marini, NCSA