Elli Angelopoulou is an admitted chocoholic. She is also a doctoral student in computer science at Johns Hopkins. So it was natural that her interest in computer vision led to the development of a system that can, among other things, tell the difference between two homemade chocolates (upper left and right). Her hardware is strictly off the shelf: a desktop computer, three lightbulbs, and a black-and-white video camera. But thanks to a program she’s written, it can do what other computer recognition systems can’t: discriminate among subtly different rounded objects. Most computer vision methods depend on the detecting of edges, lines, and corners. None of these systems are applicable when we have a smooth and free-form object like a teapot, a cup, or a mug, says Angelopoulou. Her program tracks the intensity of light reflected off an object illuminated from three different directions and uses this information to measure curved surfaces. In the computer images, red and yellow areas are the most sharply curved, blue the least. (The blue fringe around the image at lower right shows chocolate shavings on the plate the candy was on.) Angelopoulou says robots could use her program to sort through toys or candies.