Overall Comments On Images

The stool images were acquired by taking videos on a handheld Sony camcorder; all other images were acquired by taking photographs on an Olympus digital camera.  All images were lit by fluorescent light, except the bike set, which was lit by a combination of fluorescent and natural light.  The target objects, clutter objects, and camera rotate and translate between images; however, we did not carefully measure exact 3D poses for any object in any image since the goal of these data sets was to quickly acquire large sets of images which overall span a wide range of object and scene variation.  Scale of the target object does not vary significantly in any images except those of the stool.  We detected binary edges in the scene using the Vista line finder; edge detection parameters, where appropriate, are listed.