294878.fig.0010a
(a)
294878.fig.0010b
(b)
294878.fig.0010c
(c)
294878.fig.0010d
(d)
294878.fig.0010e
(e)
294878.fig.0010f
(f)
Figure 10: Error for each of the networks’ O, C, and OBCG layers when a partial context and full object were given as input. (a) SplitDG, (b) SplitDG+, (c) AllSplit, (d) AllSplit+, (e) Baseline, and (f) average error across the object output and the lowest of the two context outputs (C or OBCG) for each network, as a function of percentage of context input presented.