The availability of a large, freely redistributable set of highquality annotated images is critical to allowing researchers in the area of automatic annotation, generic object recognition and concept detection to compare results. The recent
Jonathon S. Hare, Paul H. Lewis