Table 13.

Comparisons of recall on RMBD-DB1.

Methods	IoU Threshold														Average
	0.3	0.35	0.4	0.45	0.5	0.55	0.6	0.65	0.7	0.75	0.8	0.85	0.9	0.95
YOLOv2 [45]	0.959	0.947	0.928	0.898	0.851	0.749	0.596	0.410	0.241	0.129	0.050	0.019	0.003	0.000	0.484
DeepDeblur [23] + YOLOv2	0.991	0.986	0.978	0.956	0.907	0.799	0.657	0.459	0.264	0.127	0.057	0.016	0.005	0.001	0.514
DeblurGAN [22] + YOLOv2	0.960	0.949	0.933	0.907	0.871	0.829	0.756	0.657	0.534	0.376	0.203	0.080	0.021	0.002	0.577
DeblurGAN (MobileNet) [50] + YOLOv2	0.973	0.966	0.957	0.942	0.915	0.861	0.798	0.695	0.540	0.367	0.176	0.075	0.018	0.003	0.592
DCSCN + lightDenseYOLO [21]	0.960	0.947	0.929	0.899	0.852	0.749	0.597	0.409	0.242	0.128	0.051	0.019	0.004	0.000	0.485
SlimDeblurGAN + YOLOv2 (ours)	0.975	0.969	0.955	0.935	0.911	0.870	0.810	0.709	0.570	0.411	0.220	0.103	0.032	0.005	0.605