This was my final project for
Computational Photography (Brown CS129). It applied established algorithms for image
quality assessment at a novel scale to estimate empirically the effect of training database size. The features for the
images are derived from Y. Ke, et. al. (CVPR '06) and reflect common compositional attributes of “good” photos
(Ke et. al. attempt to distinguish between professional and amateur photographs). This project uses a new dataset
derived from approximately one year's worth of Flickr data, using the most and least “interesting” photos from
each day, amounting to about 1.4 million photographs.
See »