505191.fig.002
Figure 2: Learning of motion models and semantic knowledge.