Visual Speech Recognition

Benchmarks

Dataset Name	SOTA Method	Metric
LRS3 (test)	Llama-AVSR	WER0.77	209	1mo ago
LRS3 High-Resource, 433h labelled v1 (test)		WER0.009	80	1mo ago
LRS3	Auto-AVSR	WER0.009	63	25d ago
LRS2	AutoAVSR	Mean WER14.6	49	25d ago
LRS3 Low-Resource 30h labelled v1 (test)	USR	WER0.024	34	1mo ago
LRS3 30h labeled low-resource (test)	UASR-LLM-L	WER25.3	28	1mo ago
LRS3 low-resource (test)		WER19.3	20	1mo ago
LRS3 high-resource (test)	RAVEn w/ self-training	WER23.1	16	1mo ago
WildVSR	Auto-AVSR	WER38.6	15	1mo ago
LRS2 v0.4 (test)	Ours (raw A + V)	WER3.7	14	1mo ago
CMLR (Seen)	Cascade-Free Mandarin VSR	CER20.38	12	25d ago
LSVSR (test)	Audio-Ph	Word Error Rate18.3	10	1mo ago
LRS3 v0.4 (test)	Ours (raw A + V)	WER2.3	9	1mo ago
CMLR (Unseen)	Cascade-Free Mandarin VSR	CER38.23	8	25d ago
CMLR (test)	Lipnet	Inference Latency (ms)52.3	8	25d ago
CMLR	VSR model with prediction-based auxiliary tasks	Best CER8	7	1mo ago
CNVSRC-Multi Mandarin (dev)	VALLR-Pin	CER24.1	6	1mo ago
LRW	SyncVSR	Top-1 Accuracy80.3	5	1mo ago
LRS3 v0.0 (test)	Ours (raw A + V)	WER1.2	5	1mo ago
Self-Collected Dataset Mandarin (test)	VALLR-Pin	CER32.22	4	1mo ago
CMU-MOSEAS-Spanish (CMes)	CM-seq2seq	Best Score58.1	4	1mo ago
CMU-MOSEAS-Portuguese (CMpt) (test)	VSR model with prediction-based auxiliary tasks	Mean WER51.6	4	1mo ago
Multilingual TEDx-Spanish (MTes) (test)	VSR model with prediction-based auxiliary tasks	Mean WER56.6	4	1mo ago
CMU-MOSEAS French	VSR model with prediction-based auxiliary tasks	Mean WER59.1	4	1mo ago
Multilingual TEDx-Portuguese (MTpt) (test)	CM-seq2seq	Mean Accuracy70.2	4	1mo ago

Showing 25 of 28 rows