List your company
ResearchFormerly WizWand

Visual Speech Recognition on LRS3 high-resource (test)

23.1WERRAVEn w/ self-training
20.53237.86655.272.534Dec 12, 2022
Updated 8mo ago

Evaluation results

MethodLinks
RAVEn w/ self-trainingParadigm=self-supervised (large, more pre-training), Encoder=Transformer, Language Model=true, Unlabeled hrs=1759, Labeled hrs=4332022.1223.1
RAVEn w/ self-trainingParadigm=self-supervised (large, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1224.4
Shi et al. (2022) w/ self-trainingParadigm=self-supervised (large, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1226.9
RAVEnParadigm=self-supervised (large, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1227.8
Shi et al. (2022)Paradigm=self-supervised (large, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1228.6
Ma et al. (2022)Paradigm=semi-supervised, Encoder=Conformer, Language Model=true, Unlabeled hrs=641, Labeled hrs=8182022.1231.5
RAVEnParadigm=self-supervised (base, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1233.1
Shi et al. (2022)Paradigm=self-supervised (base, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1234.8
Large model w/ curriculumParadigm=from scratch, Encoder=Transformer, Language Model=false, Labeled hrs=4332022.1238.8
RAVEnParadigm=self-supervised (base, less pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=433, Labeled hrs=4332022.1239.1
Base model w/ curriculumParadigm=from scratch, Encoder=Transformer, Language Model=false, Labeled hrs=4332022.1239.8
Shi et al. (2022)Paradigm=self-supervised (base, less pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=433, Labeled hrs=4332022.1244
Ma et al. (2021a)Paradigm=self-supervised (base, more pre-training), Encoder=Transformer, Language Model=false, Unlabeled hrs=1759, Labeled hrs=4332022.1249.6
Afouras et al. (2020)Paradigm=semi-supervised, Encoder=CNN, Language Model=true, Unlabeled hrs=344, Labeled hrs=4332022.1259.8
Large modelParadigm=from scratch, Encoder=Transformer, Language Model=false, Labeled hrs=4332022.1285.8
Base modelParadigm=from scratch, Encoder=Transformer, Language Model=false, Labeled hrs=4332022.1287.3