Speech Enhancement on LRS2 mixed with VGGSound noises (test) benchmark leaderboard · SOTA2 Research

Speech Enhancement on LRS2 mixed with VGGSound noises (test)

3.22PESQVisual-stream method [2]

Evaluation Results

Method
Visual-stream method [2]2020.12	3.22	3.49	2.73	2.39	0.93
pseudo-visual speech enhancement2020.12	3.19	3.44	2.71	2.35	0.92
AO (Audio-Only)2020.12	3.1	3.36	2.63	2.27	0.89
Visual-stream method [2]2020.12	3.04	3.38	2.63	2.31	0.9
pseudo-visual speech enhancement2020.12	2.98	3.32	2.57	2.26	0.9
AO (Audio-Only)2020.12	2.87	3.21	2.48	2.17	0.88
Visual-stream method [2]2020.12	2.79	3.23	2.47	2.2	0.89
Method [11]2020.12	2.79	3.33	2.57	2.23	0.89
pseudo-visual speech enhancement2020.12	2.71	3.16	2.41	2.15	0.87
Noisy2020.12	2.69	3.23	2.58	2.2	0.85
AO (Audio-Only)2020.12	2.58	3.07	2.31	2.07	0.84
Method [11]2020.12	2.55	3.16	2.42	2.14	0.85
Method [19]2020.12	2.41	3.13	2.59	2.11	0.86
Noisy2020.12	2.32	3.01	2.28	2.16	0.84
Method [11]2020.12	2.1	2.8	2.22	1.93	0.8
Noisy2020.12	1.94	2.55	1.86	1.81	0.75
Method [19]2020.12	1.87	2.85	2.12	2.03	0.83
Method [19]2020.12	1.82	2.29	1.82	1.59	0.73