Public Eyesores Disappear With Camouflage Tech
MIT researchers create an algorithm to cloak public eyesores like electrical boxes.
Eyesores, beware! A team of researchers recently unveiled a computer algorithm that can use photos from a scene to create a covering for eyesores in otherwise beautiful places like historic landmarks or picturesque trails. The algorithm can also be used to cover objects in ordinary areas, like electrical boxes outside of homes. MIT researchers teamed up with other institutions to create the eyesore algorithm solution, revealed at a conference on Computer Vision and Pattern Recognition in June.
The computerized algorithm analyzes photos of an area and creates color variations for eyesores based on the colors and patterns in the surrounding area. The algorithms vary in complexity, with the simplest using the average color values to produce a uniform hue to cover all surfaces. The more nuanced camouflage covers blend more into the surroundings than a uniform cover that may stand out from certain angles.
Researchers tested their methods using Amazonâ€™s Mechanical Turk crowdsourcing application. Each algorithm was scored based on the amount of time people took to locate camouflaged objects in synthetic images. The best performing algorithm, the “greedy” algorithm, had people looking for over three seconds to find the camouflaged item.
The greedy algorithm works by creating the illusion of circular objects that blend into the surroundings. It identifies the camera angles that require the least distortion of the patterns applied to each face of an object.
Researchers are still working to fine-tune the algorithms for real-life circumstances such as changing light conditions, but the general methodology of the camouflage creator has set the stage for a new breakthrough in cityscaping.
Andrew Owens from MIT, William Freeman from MIT, Connelly Barnes from the University of Virginia, Alex Flint from Flyby Media and Hanumant Singh from Woods Hole Oceanographic Institution worked on the computer-assisted camouflage research project.
[h/t] Custom Camouflage