Face Tracking with Visual Attention (SD)

PittPatt has developed state-of-the-art face tracking capabilities. This is a demonstration of visual attention applied to face tracking. Visual attention is meant as an extremely fast first-pass at determining which regions of a frame should be analyzed. The bright portions show which parts of each frame are searched for faces. The dark regions represent the parts that were skipped (and thus, represent computational savings). The skipped regions have their results propagated forward from the previous frame. The algorithm periodically searches the entire frame to ensure that small error does not accumulate over time. The frame rate has been intentionally slowed down. See http://www.pittpatt.com for more information.

Category:

Science & Technology

Tags:

• pittsburgh
• pattern
• recognition
• pattern recognition
• pittpatt
• computer vision
• face detection
• face tracking
• visual attention