FIGURE 12.

Example frames captured by the eye camera that are presented as visual input to the trained models with their subsequent reconstructions in the 2 sensory drop-out conditions. The Ground truth images in the left column were taken at the sample number indicated to the left of each panel with each reconstruction from the models presented in that row. The reconstructions are qualitatively similar in their quality across all 3 models, however, MultiPredNet did return the lowest mean squared reconstruction errors in all conditions as shown in Figure 8.