A Recurrent Temporal Model for Semantic Levels Categorization Based on Human Visual System

Figure 14

The occluded images for the ten categories. These images are shown with 40 to 60 occlusion percent, which are created randomly and used to test the effect of feedback in the case of occlusion.