Feature Guided CNN for Baby’s Facial Expression Recognition

Table 4

The expression recognition performance of different methods on the BabyExp dataset (trained from scratch).

Models	Happy			Sad			Normal			Average accuracy
	Recall	Precision	F1	Recall	Precision	F1	Recall	Precision	F1
DLP [31]	62.25	66.76	64.42	53.04	82.96	64.90	79.70	56.20	65.92	65.02
Baseline [25]	65.13	80.53	72.01	82.24	88.86	85.42	91.21	72.18	80.59	79.57
SO-DL	52.75	92.34	67.14	93.91	85.52	89.52	97.03	73.13	83.40	81.31
SO-DLSE	59.62	91.73	72.27	93.54	89.22	91.33	96.04	73.98	83.58	83.13
VFESO-DLSE	78.5	93.18	85.21	96.27	85.07	90.33	88.86	86.71	87.78	87.90