Learning visual composition preferences from an annotated corpus generated through gameplay | IEEE Conference Publication | IEEE Xplore