Deep Ensemble Learning for Human Action Recognition in Still Images

Table 8

Results for DELVS in the Willow action dataset.

Model	Strategy	Sensitivity for each class							Overall
		Inter.W.C.	Photog.	P.Music	R.Bike	R.Horse	Running	Walking	Acc
DELVS1	Hard	0.9	0.57	0.75	0.97	0.88	0.52	0.22	0.6727
	Soft	0.9	0.61	0.78	0.97	0.88	0.6	0.37	0.7189
	Tuning	0.93	0.6	0.7890	0.96	0.88	0.59	0.4	0.7229
DELVS2	Hard	0.83	0.53	0.75	0.93	0.92	0.64	0.37	0.6968
	Soft	0.9	0.48	0.73	0.95	0.92	0.62	0.45	0.7048
	Tuning	0.9	0.45	0.75	0.95	0.94	0.64	0.51	0.7189
DELVS3	Hard	0.9	0.59	0.75	0.95	0.94	0.62	0.32	0.7048
	Soft	0.93	0.56	0.76	0.95	0.92	0.62	0.41	0.7189
	Tuning	0.9	0.65	0.79	0.95	0.94	0.64	0.39	0.7369