Instructor Activity Recognition through Deep Spatiotemporal Features and Feedforward Extreme Learning Machines

Table 3

Performance of the proposed technique at different network depth on the IAVID-I dataset using 70-30 validation scheme.

Deep spatiotemporal features	Features	No. Layers	Hidden Node	Accuracy %	Time; Sec
x17	1x 4096	17	1300	81.43	0.00159
x 20	1x 4096	20	1300	78.34	0.00203
x 23	1x 1000	23	1300	73.98	0.00108
x 39	1x 4096	39	500	78.23	0.00059
x 42	1x 4096	42	500	75.88	0.001302
x 45	1x 1000	45	500	74.09	0.000811