Table 1.

SD-resolution comparison experiments on the VM dataset.

Model	MAD (1 × 103)	MSE (1 × 103)	Grad	Conn	dtSSD (1 × 102)
DeepLabV3	14.47	9.67	8.55	1.69	5.18
FBA	8.36	3.37	2.09	0.75	2.09
BGMv2	25.19	19.63	2.28	3.26	2.74
MODNet	9.41	4.30	1.89	0.81	2.23
VideoMatte	6.06	1.27	1.09	0.42	1.60
RVM	6.08	1.47	0.88	0.41	1.36
RVM-Large	5.66	0.92	0.75	0.37	1.32
Our Model	5.39	0.88	0.69	0.28	1.17
Our Model + CE	5.31	0.79	0.65	0.26	1.08