Table 4.

Results of the action recognition of the College dataset.

Video-AP (%)	Normal	Abnormal	mAP
Novel Spatiotemporal Network	68	61	64.5
NEUCOM21	71	52	61.5
Improved Robust Video Saliency Detection based on Long-term Spatial-temporal Information	69	51	60
BDI-CNN	75.8	51.9	63.85
MHI-CNN	69.6	61.3	65.4
WAI-CNN	79.9	74.1	77
OFSDI	82.1	78.4	80.25