Video wird geladen...
Video konnte nicht geladen werden
ð£ïž Speech-to-Text with Lip Reading ð¬ æ³šç®ããŠãã人ã®å£°ã ããèãåãããã«ãã¢ãŒãã«é³å£°èªèã¢ãã«ãAVistaãããªãªãŒã¹ããŸããïŒ ãã¥ãŒããã€ããçæŽ»ã®å Žã§æŽ»èºããã«ã¯ãå°ãé¢ããå Žæããã®åŒã³ãããæ£ç¢ºã«èãåãå¿ èŠããããŸããåŸæ¥ã®é³å£°èªèã¯ãåæçºè©±ãåšå²ã®éé³ãå¢ãããšç²ŸåºŠãèœã¡ãããåŸåããããŸããã AVistaã¯é³å£°ã«åã®åãïŒæ åïŒãçµã¿åãããããšã§ããã®ãããªæ··ã¿åã£ãç°å¢ã§ããæ³šç®ããŠãã人ã®å£°ãéžæçã«æåèµ·ããã§ããŸãã AVista: Audio-Visual Transcription and Alignment ð
26,697 Aufrufe ⢠vor 10 Monaten â¢via X (Twitter)
7 Kommentare

AVistaã«ã¯AVHuBERTãšããææ³ãæ¡çšããŸããããã®ææ³ã¯ãé³å£°ãŸãã¯åç»ã®ãã¹ã¯ãããé åã®ãŠããããäºæž¬ããSSLãã¬ãŒã ã¯ãŒã¯ã§ãã ä»åã®ãªãªãŒã¹ã§ã¯ãæ¥æ¬èªã«ç¹åããè€æ°ã®äºååŠç¿æžã¿ã¢ãã«ããã³AVSRã¢ãã«ãåæã«å ¬éããŠããŸããTransformersã©ã€ãã©ãªã䜿çšããããã«ã¢ãã«ã®æšè«ã詊ãããšãã§ããŸãïŒ

@enactic_ai, AVista's multimodal approach to speech recognition is quite interesting, especially for noisy environments.

åé¢ããŠå šå¡åèŽãã®æ¬²ãããŒãŒ

@ReazonHILab Congratulations team Reazon HI team. This is huge step forward in Japanese ASR. Good job!

nice- for robots or for the fbi

Cool.

Quite the visual. A bit too metaphysical for my usual feed though.
