Encoding Longer-Term Contextual Information with Predictive Coding and Ego-Motion

Figure 17

Representation in GU Units (Layer 2) (Crazy turn).

(a) Frame 10 (GU 1)

(b) Frame 50 (GU 1)

(c) Frame 100 (GU 1)

(d) Frame 150 (GU 1)

(e) Frame 200 (GU 1)

(f) Frame 10 (GU 2)

(g) Frame 50 (GU 2)

(h) Frame 100 (GU 2)

(i) Frame 150 (GU 2)

(j) Frame 200 (GU 2)