List your company
ResearchFormerly WizWand

Word Recognition on LRW (test)

98Correct RateA + V (End-to-End)
59.62469.58779.5589.513Feb 18, 2018May 22, 2018Aug 23, 2018Nov 24, 2018Feb 25, 2019May 29, 2019Aug 30, 2019
Updated 8mo ago

Evaluation results

MethodLinks
A + V (End-to-End)Stream=Audiovisual, Model Type=End-to-End, Mouth ROI=Fixed2018.0298
A (End-to-End)Stream=Audio, Model Type=End-to-End, Input=Raw waveform2018.0297.7
A (MFCC)Stream=Audio, Model Type=2-layer BGRU, Features=MFCC2018.0297.7
Multi-Grained Spatio-temporal Network2019.0883.34
V [13]*Stream=Video, Model Type=End-to-End, Mouth ROI=Tracked facial landmarks2018.0283
Stafylakis172019.0883
V (End-to-End)Stream=Video, Model Type=End-to-End, Mouth ROI=Fixed2018.0282
Petridis18mode=end-to-end2019.0882
Petridis18reproduction=true2019.0881.7
V [15]Stream=Video2018.0276.2
Chung172019.0876.2
Chung182019.0871.5
V [19]Stream=Video2018.0261.1