COSMOS researchers earned the “Best Paper Award” for their submission “YouTube Video Categorization Using Moviebarcode” at the International Academy, Research, and Industry Association (IARIA) Conference on Human and Social Analytics (HUSO 2020), held in Porto, Portugal. 

The paper written by Recep Erol, Rick Rejeleene, Richard Young, Thomas Marcoux, Dr. Muhammad Nihal Hussain, and Dr. Nitin Agarwal introduces the idea of utilizing moviebarcode, a technique used to summarize videos by compressing an entire video into a single image, to systematically categorize YouTube videos. “Moviebarcodes are typically used to visualize summary of videos,” stated lead author Recep Erol. “However, we used the color theory computationally so that we can categorize videos using this technology.” 

The idea stemmed from a project Richard Young worked on in the Spring of 2019 for the Social Computing course offered by Dr. Agarwal. “What I find interesting about movie barcodes is that hours of video content can be quickly digested by a user in a single glance,” Young stated. “After researching moviebarcodes further, I began considering possible applications to my work with COSMOS proposing the development of a moviebarcodes tool that can effectively visualize video data and provide impactful insights.”

Each barcode consists of generated colors for every frame of the movie making it unique. All frames combined into one code showed color transition within videos and allows for a comparison with other videos and a highly accurate grouping of videos without having to watch them. 

The researchers analyzed videos of different length, which presented a challenge. Longer videos first generated longer moviebarcodes.  “In order to overcome this problem, I put moviebarcodes in a standard scale so that I could summarize videos regardless of their length,” explained Erol. He was also able to reduce time and computing energy required to categorize videos by creating video collections and then categorizing videos using a clustering method. “This approach shortens the waiting time for data researchers and saves computational energy,” Erol pointed out.

Dr. Agarwal said, “cosmographers have a knack of thinking out of the box and producing groundbreaking innovations.” He continued, “Moviebarcode based video data analysis is one such idea with numerous applications that COSMOS is developing including faster and efficient search and retrieval, computationally extracting noteworthy narrative elements in videos, identifying coordinated influence campaigns, among others.”

Best paper decisions consider reviewers’ feedback, discussion during the conference presentation, presentation style, and post-conference feedback from participants. The paper presentation is available here.

This research is funded in part by the U.S. Air Force Research Lab, U.S. Office of Naval Research, U.S. Army Research Office, U.S. Defense Advanced Research Projects Agency, U.S. National Science Foundation, Arkansas Research Alliance, and the Jerry L. Maulden/Entergy Endowment at the University of Arkansas at Little Rock. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the funding organizations. The researchers gratefully acknowledge the support.