KAIST Develops Lip-Reading AI That Decodes Clear Words First to Resolve Ambiguous Mouth Movements (IMAGE)
Caption
Conceptual comparison between autoregressive decoding and the proposed method. (a) The conventional approach generates words one at a time from left to right, so it cannot revise an early error, such as misrecognizing “PACK” as “BACK.” (b) The proposed method first identifies words it can recognize with confidence, such as “BAGS” and “TRIP,” and then uses them as context to correctly identify the ambiguous early word.
Credit
KAIST
Usage Restrictions
No
License
Licensed content