Facial Expression Recognition of Instructor Using Deep Features and Extreme Learning Machine

Table 2

Statistical performance of seven 2D-CNN models (AlexNet, DenseNet201, ReseNet18, ResNet50, ResNet100, GoogleNet, and Inceptionv3) on the Instructor Expression Video (IEV).

Network	Layer	Accuracy	Neurons	Precision	Recall	F1-score	Error
DenseNet201	conv4_block9_1_bn	0.85	5000	0.85	0.87	0.85	0.14
AlexNet	drop7	0.83	5000	0.83	0.85	0.83	0.12
GoogleNet	pool5-drop_7 × 7_s1	0.81	5000	0.81	0.83	0.81	0.18
Inceptionv3	activation_94_relu	0.78	5000	0.75	0.79	0.76	0.22
ResNet101	pool5	0.84	5000	0.83	0.85	0.83	0.16
ResNet18	pool5	0.75	5000	0.79	0.81	0.79	0.2
ResNet50	avg_pool	0.82	5000	0.82	0.82	0.82	0.82