Table 2.

Classification performance.

Model	Rest Classifier			Stress Classifier			Trainable Parameters
	Agreement	Sensitivity	Specificity	Agreement	Sensitivity	Specificity
FCN	84.3	80.5	87.0	83.3	71.4	93.8	212,982
CNN	83.7	76.8	90.0	82.9	65.8	98.0	12,462,594
GCNN Chebyshev	89.3	85.1	92.8	91.2	85.7	95.9	12,399,730
GCNN Cayley	88.8	86.9	90.0	90.8	84.3	95.8	5,9999,204

Best results are shown in bold and all values are in percentage of agreement with the human observer.