From: Human detection in surveillance videos and its applications - a review
Original frame
Ground truth
MoG
NP-KDE
Temporal median
Frame difference
Output
FP
21.5%
2.3%
6.9%
FN
12.9%
11.5%
8%
13.4%