Abstract: We present a systematic study of generative data for Japanese sentence-level lip-reading, demonstrating that synthetic speech videos can substantially improve accuracy in a lowresource ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results