Method DMSN (SIE) on the SBM-RGBD dataset

Contact name: Ihssane Houhou

Contact email: ihssane.houhou@univ-biskra.dz

Contact Institution: University of Biskra, Algeria / University of Technology Belfort-Montbeliard, France

Method processing time: ~4 fps using Python 3.6, Keras 2.3, Tensorflow-gpu 2.1, and Cuda 10.1 installed on a machine built with Intel i9 CPU 9th generation and GeForce RTX 2080 GPU under Windows 10 operating system.

Link to the code: https://github.com/ihssanehouhou/DMSN

Reference: I. Houhou, A. Zitouni, Y. Ruichek, SE. Bekhouche, M. Kas, A. Taleb-Ahmed, RGBD deep multi-scale network for background subtraction, International Journal of Multimedia Information Retrieval, vol.11, pp.395-407, May, 2022.

Note: DMSN (SIE) results were obtained by using a different protocol (please check the referenced paper).

Get all result files: here.

A) Sequence-by-sequence results

Name	Recall	Specificity	FPR	FNR	PWC	Precision	F-Measure
Bootstrapping
BootStrapping_ds	0.6596	0.9968	0.0032	0.3404	3.4901	0.9549	0.7803
adl24cam0	0.7798	0.9998	0.0002	0.2202	1.3541	0.9961	0.8748
bear_front	0.4778	0.9265	0.0735	0.5222	19.1692	0.6991	0.5676
fall01cam0	0.8954	0.9920	0.0080	0.1046	1.3322	0.8688	0.8819
fall20cam0	0.8464	0.9589	0.0411	0.1536	4.6858	0.5281	0.6504
ColorCamouflage
Cespatx_ds	0.5867	0.9994	0.0006	0.4133	0.9021	0.9540	0.7266
Hallway	0.3006	0.9938	0.0062	0.6994	4.2426	0.7275	0.4254
colorCam1	0.0000	1.0000	0.0000	1.0000	22.3980	0.4091	0.0001
colorCam2	0.0682	0.9997	0.0003	0.9318	42.2793	0.9940	0.1276
DepthCamouflage
DCamSeq1	0.8610	0.9936	0.0064	0.1390	1.3841	0.8877	0.8741
DCamSeq2	0.8047	0.9989	0.0011	0.1953	0.9604	0.9707	0.8800
Despatx_ds	0.9378	0.9900	0.0100	0.0622	1.4738	0.9034	0.9203
Wall	0.1569	0.9993	0.0007	0.8431	7.5907	0.9580	0.2696
IlluminationChanges
ChairBox	0.6589	0.9845	0.0155	0.3411	5.0953	0.8387	0.7380
Ls_ds	0.0000	0.9999	0.0001	0.0000	0.0107	0.0000	0.0000
TimeOfDay_ds	0.0000	0.9988	0.0012	0.0000	0.1223	0.0000	0.0000
genSeq1	0.9841	0.9961	0.0039	0.0159	0.5447	0.9730	0.9785
IntermittentMotion
Shelves	0.2864	0.9945	0.0055	0.7136	1.9347	0.5107	0.3670
Sleeping_ds	0.8446	0.9935	0.0065	0.1554	3.6378	0.9705	0.9032
abandoned1	0.9804	0.9999	0.0001	0.0196	0.0583	0.9963	0.9883
abandoned2	0.9616	0.9988	0.0012	0.0384	0.5738	0.9908	0.9760
movedBackground1	0.8938	0.9999	0.0001	0.1062	0.1642	0.9946	0.9415
movedBackground2	0.9426	0.9968	0.0032	0.0574	0.6695	0.9523	0.9474
OutOfRange
MultiPeople1	0.9167	0.9927	0.0073	0.0833	1.3252	0.9142	0.9154
MultiPeople2	0.9511	0.9882	0.0118	0.0489	1.4827	0.8774	0.9128
TopViewLab1	0.9508	0.9984	0.0016	0.0492	0.3156	0.9542	0.9525
TopViewLab2	0.9660	0.9987	0.0013	0.0340	0.2555	0.9675	0.9668
TopViewLab3	0.8402	0.9990	0.0010	0.1598	0.4420	0.9460	0.8900
Shadows
Shadows_ds	0.7498	0.9974	0.0026	0.2502	1.7787	0.9503	0.8383
fall01cam1	0.8783	0.9400	0.0600	0.1217	6.3918	0.4958	0.6338
genSeq2	0.8711	0.9985	0.0015	0.1289	1.5747	0.9861	0.9250
shadows1	0.9584	0.9984	0.0016	0.0416	0.7186	0.9900	0.9740
shadows2	0.5860	0.9996	0.0004	0.4140	5.7083	0.9952	0.7376

B) Average results for each category

Name	Recall	Specificity	FPR	FNR	PWC	Precision	F-Measure
Bootstrapping	0.7318	0.9748	0.0252	0.2682	6.0063	0.8094	0.7510
ColorCamouflage	0.2389	0.9982	0.0018	0.7611	17.4555	0.7712	0.3199
DepthCamouflage	0.6901	0.9954	0.0046	0.3099	2.8522	0.9300	0.7360
IlluminationChanges	0.4107	0.9948	0.0052	0.0893	1.4433	0.4529	0.4291
IntermittentMotion	0.8182	0.9972	0.0028	0.1818	1.1730	0.9025	0.8539
OutOfRange	0.9250	0.9954	0.0046	0.0750	0.7642	0.9319	0.9275
Shadows	0.8087	0.9868	0.0132	0.1913	3.2344	0.8835	0.8217
AVG.	0.6605	0.9918	0.0082	0.2681	4.7041	0.8116	0.6913