Anomaly Detection in Videos Using Two-Stream Autoencoder with Post Hoc Interpretability

Table 4

Comparison of the average computation time (per epoch) on four data sets.

Method	Avenue (m)	Ped2 (m)	Subway Entrance (m)	Subway Exit (m)
ST-AE [20]	180	30	640	360
ConvAe [7]	244	40	320	120
Conv-LSTM-AE [18]	312	73	766	452
GMFC-VAE [11]	220	65	712	432
DAEs + cGAN [14]	430	194	904	642
Spatial model	19	3	24	8
Temporal model	95	15	120	50
Spatial + temporal	142	24	182	70