Table 1.

Performance comparison of AlexNet, ResNet101, and Inception-ResNet-v2 in terms of decoding accuracies with spatial-spectral-temporal features.

CNN architecture	Pre-stimuli (%)	Perception (%)	Preparation (%)	Production (%)
AlexNet	43.77	90.38	93.24	96.65
ResNet101	42.61	84.75	87.78	92.36
Inception-ResNet-v2	42.52	87.98	91.66	94.49
Mean	42.96	87.70	90.89	94.50
STD	0.70	2.82	2.81	2.14

The test accuracies are averaged over 3 independent runs across subjects.