Fusion Attention Mechanism for Foreground Detection Based on Multiscale U-Net Architecture

Table 6

F-measure comparison of different methods on CDnet-2014 dataset.

Method	PTZ	badWeat	Baseline	cameraJit	dynaBg	Intermit	lowFrame	nightVid	Shadow	Thermal	Turbul	Overall
AMU-Net_M1	0.9684	0.9865	0.9936	0.9884	0.9909	0.9897	0.9603	0.9733	0.9928	0.9898	0.9764	0.9827
AMU-Net_M2	0.9759	0.9829	0.9941	0.9873	0.9913	0.9888	0.9596	0.9741	0.9930	0.9900	0.9754	0.9830
AMU-Net	0.9759	0.9875	0.9936	0.9913	0.9914	0.9879	0.9030	0.9779	0.9921	0.9880	0.9751	0.9785
CascadeCNN [28]	0.9344	0.9451	0.9786	0.9758	0.9658	0.8505	0.8804	0.8926	0.9593	0.8958	0.9215	0.9272
SuBSENSE [21]	0.3476	0.8619	0.9503	0.8152	0.8177	0.6569	0.6445	0.5599	0.8986	0.8171	0.7792	0.7408
FTSG [23]	0.3241	0.8228	0.9330	0.7513	0.8792	0.7891	0.6259	0.5130	0.8535	0.7768	0.7127	0.7283
GMM [14]	0.1046	0.7406	0.8382	0.5670	0.6328	0.5325	0.5065	0.3960	0.7322	0.6548	0.4169	0.5566