A new method developed by Disney Research scientists should enable automated cameras to produce video imagery that feels less robotic but is smooth and aesthetically pleasing. Instead of having the automated system track a key object, it would learn from a human operator how to better frame shots.
Disney Research engineer Peter Carr and University of British Columbia Ph.D. student Jianhui Chen demonstrated their data-driven approach during a high school basketball game. The system monitors a camera operator during the game, and machine-learning algorithms enable it to recognize the relationship between player locations and corresponding camera configurations.
"We don't use any direct information about the ball's location because tracking the ball with a single camera is difficult," Carr notes. "But players are coached to be in the right place at the right time, so their formations usually give strong clues about the ball's location."
The researchers report the system mimicked the human operator of the broadcast camera, and the system's prediction for panning the camera was superior to the best previous algorithms.
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found