Home Research Research Highlights
NIMBLE Eyes
Researchers at the NSF-sponsored Temporal Dynamics Learning Center, located at the University of California, San Diego, have recently developed a computer system called NIMBLE. The system recognizes faces and objects in a manner similar to how people preform the same task: by strategically moving its “eyes” across the face or object. The system has a built-in way of deciding what is “interesting” in an image, and “looks” there. Gary Cottrell and his graduate student Luke Barrington, along with postdoctoral researchers Tim Marks and Janet Hsiao, built the system to better understand how humans use their eyes to gather information from their visual environment. By comparing the system’s eye movements to human subjects’ eye movements over the same faces, they hope to make the system more and more human-like. Such a system can be much less sensitive to changes in lighting, head orientation, and position because it first seeks out the interesting places on the face and then examines “patches” of the image from that location. Small patches of an image are less subject to variation under these alternate lightings and viewpoints. Such a system may someday be used to recognize faces in the same conditions that we do.