At the upcoming Computer Vision and Pattern Recognition conference, researchers from MIT and other institutions will be figuring out how to hide obtrusive objects in public spaces by using new algorithms. The goal is to camouflage items such as electrical boxes and portable toilets into their surrounding environments. The algorithm itself would work by analyzing photos of a scene from multiple perspectives, and then produce a way to blend the object into the area with a camouflage covering.
These algorithms have already been tested using Amazon's Mechanical Turk crowdsourcing application. They tested to see if volunteers could find the camouflaged objects in synthetic images. The best performers took an average of more than three seconds to find the objects, which is much longer than the usual casual glance.
"Usually these algorithms exploit certain cues -- maybe they're looking for the contours of the object, or the silhouette of the object, the boundaries," Andrew Owens, an MIT graduate and lead author of the new paper, told PhysOrg. "With camouflage, you want to avoid these cues. Conceptually, a cue that would be good for detecting an object is something that you want to remove."
The algorithms begin at the point in which one angle produces a perfectly blended image, then makes tradeoffs based on other angles to ensure it is always blended.
"This is a new task, which makes it fun," says James Hays, an assistant professor of computer science at Brown University. "They use texture-synthesis techniques that are pretty well known in the community, but they've never been used in this way. They've been used for things like photo editing, but never for designing a real-world artifact. So that's innovative."
Hays notes that the experiments conducted involve the images taken under just about the same lighting conditions. However, in the real world, a working camouflage covering would need to accommodate for different times of day.