Table 2.

AU detection performances on MIAMI dataset

−	S	AUC	PA	NA
(a) AU4
Study 1: MIAMI $\to$ MIAMI	0.88	0.83	0.69	0.97
Study 2: (MIAMI + CLOCK) $\to$ MIAMI	0.87	0.80	0.65	0.96
Study 3: CLOCK $\to$ MIAMI	0.80	0.79	0.57	0.94
OpenFace	−0.08	0.64	0.25	0.58
Study 4 - Adult AFAR: (EB+ + GFT) $\to$ MIAMI	0.74	0.82	0.54	0.92
Study 5 - Infant AFAR: (EB+ + GFT + MIAMI + CLOCK) $\to$ MIAMI	0.90	0.85	0.73	0.97
(b) AU6
Study 1: MIAMI $\to$ MIAMI	0.75	0.86	0.81	0.90
Study 2: (MIAMI + CLOCK) $\to$ MIAMI	0.73	0.86	0.80	0.89
Study 3: CLOCK $\to$ MIAMI	0.34	0.75	0.65	0.68
OpenFace	0.51	0.63	0.43	0.85
Study 4 - Adult AFAR: (EB+ + GFT) $\to$ MIAMI	0.59	0.82	0.73	0.83
Study 5 - Infant AFAR: (EB+ + GFT + MIAMI + CLOCK) $\to$ MIAMI	0.76	0.86	0.81	0.91
(c) AU12
Study 1: MIAMI $\to$ MIAMI	0.77	0.85	0.77	0.92
Study 2: (MIAMI + CLOCK) $\to$ MIAMI	0.78	0.87	0.79	0.92
Study 3: CLOCK $\to$ MIAMI	0.72	0.81	0.72	0.91
OpenFace	0.67	0.74	0.63	0.89
Study 4 - Adult AFAR: (EB+ + GFT) $\to$ MIAMI	0.60	0.77	0.66	0.85
Study 5 - Infant AFAR: (EB+ + GFT + MIAMI + CLOCK) $\to$ MIAMI	0.76	0.84	0.77	0.91
(d) AU20
Study 1: MIAMI $\to$ MIAMI	0.81	0.79	0.64	0.94
Study 2: (MIAMI + CLOCK) $\to$ MIAMI	0.81	0.81	0.66	0.94
Study 3: CLOCK $\to$ MIAMI	0.66	0.79	0.52	0.89
OpenFace	0.56	0.55	0.21	0.87
Study 4 - Adult AFAR: (EB+ + GFT) $\to$ MIAMI	−	−	−	−
Study 5 - Infant AFAR: (EB+ + GFT + MIAMI + CLOCK) $\to$ MIAMI	−	−	−	−

Left-side of the $\to$ denotes the database(s) used to train the model in the related study. Right-side of the $\to$ denotes the database used to test the models (i.e. MIAMI)

The best results are shown in bold