go to Caltech Home
home current exhibition submission archive about us contact
    Schooner (Honorable Mention)  


next image

Greg Griffin

To train a computer to identify objects in an image, we assembled a dataset of over 30,000 images from 256 distinct object categories. Averaging all images in a single category (as shown here) shows us whether the pictures in that category are biased towards certain locations (such as the center). This is an important bias to consider if the computer is to localize objects as well as to identify them. The Caltech-256 dataset is available at here: http://www.vision.caltech.edu/Image_Datasets/Caltech256/


go to top




The 2008 Art of Science Competition was made possible by a grant from the Moore-Hufstedler Fund.
© 2018 California Institute of Technology. All Rights Reserved.