Toward Measuring Target Perception: First-Order and Second-Order Deep Network Pipeline for Classification of Fixation-Related Potentials

Figure 3

The architecture of EEGNet-, model ( and represent the number of temporal filters and spatial filters in convolutional block 1, respectively). Conv-T, Conv-S, and Conv-S-T denote the temporal, spatial, and spatiotemporal convolution, respectively. The batch normalization, average pooling, dropout, and activation layers are not depicted for clarity. The size of the output is the one from either the convolution or the activation.