Table 4.

EEGNet architecture.

Layer	Input size	Output size	Kernels	Kernel size	Stride	Padding
Conv2d	1 × 3 × 300	8 × 3 × 300	8	(1, 31)	(1, 1)	(0, 15)
BatchNorm2d	8 × 3 × 300	8 × 3 × 300
Depthwise Conv2d	8 × 3 × 300	16 × 1 × 300	16	(3, 1)	(1, 1)	(0, 0)
BatchNorm2d	16 × 1 × 300	16 × 1 × 300
Elu	16 × 1 × 300	16 × 1 × 300
AvgPool2d	16 × 1 × 300	16 × 1 × 75		(1, 4)	(1, 4)	(0, 0)
Dropout	16 × 1 × 75	16 × 1 × 75
Seperable Conv2d	16 × 1 × 75	16 × 1 × 75	16	(1, 15)	(1, 1)	(0, 7)
BatchNorm2d	16 × 1 × 75	16 × 1 × 75
Elu	16 × 1 × 75	16 × 1 × 75
AvgPool2d	16 × 1 × 75	16 × 1 × 9		(1, 8)	(1, 8)	(0, 0)
Dropout	16 × 1 × 9	16 × 1 × 9
Linear	144	2