Deep Ensemble Learning for Human Action Recognition in Still Images

Table 1

Results for NCNN in Li’s action dataset.

Algorithm	Sensitivity for each class						Overall
	Phoning	P.Guitar	R.Bike	R.Horse	Running	Shooting	Acc	Loss
SURF	0.45	0.1	0.2	0.2	0.05	0.2	0.2	NA
BOF	0.8	0.6	0.75	0.75	0.6	0.9	0.7333	NA
PBOF	0.95	0.75	0.8	0.8	0.75	0.95	0.8333	NA
GIST	0.70	0.75	0.65	0.85	0.70	0.70	0.725	NA
RF	0.70	0.75	0.55	0.75	0.85	0.70	0.7167	NA
GBM	0.55	0.55	0.75	0.70	0.65	0.60	0.6333	NA
Voting	0.85	0.75	0.80	0.85	0.75	0.70	0.7833	NA
VGG16	1	0.95	0.9	1	1	0.85	0.95	0.1564
VGG16_NCNN	1	0.9	0.95	1	1	0.85	0.95	0.1739
VGG19	0.95	0.95	0.9	1	0.8	0.85	0.9083	0.2442
VGG19_NCNN	1	0.95	1	1	0.8	0.95	0.95	0.16
ResNet50	0.9	1	0.8	0.9	1	0.95	0.925	0.2253
ResNet50_NCNN	0.95	1	0.9	1	1	0.95	0.9667	0.0703