Figure 10.

The comparison of different methods on the evaluation set. This figure illustrates the average of the required timing (y-axis) as people classifying more images (x-axis). (a): Scene8; (b): VOC2007; (c): Event Dataset; (d): Caltech101.