NVIDIA announces new platform `` Maxine'' that can process video conferencing with cloud AI



On October 5, 2020, semiconductor maker

NVIDIA announced ' Maxine ', a platform for developers that can improve the quality of video and audio of video conferencing with NVIDIA's AI and automatically process it. did.

NVIDIA Announces Cloud-AI Video-Streaming Platform to Better Connect Millions Working and Studying Remotely | NVIDIA Newsroom
https://nvidianews.nvidia.com/news/nvidia-announces-cloud-ai-video-streaming-platform-to-better-connect-millions-working-and-studying-remotely

You can see what Maxine's features are like by watching the following movie.

AI-Powered Video Conferencing with NVIDIA Maxine --YouTube


With the 'SUPER RESOLUTION' function, it is possible to improve the resolution from 360p to 720p. If you look at the patterns on women's hair and chairs, you can see that the image processed with Maxine on the right is clearer than the original image on the left.



'AUTO FRAME' is a function that automatically tracks the subject and puts it in a frame. In the following scene, a man is standing on the left side of the screen ...



The viewpoint automatically moves so that the man fits in the frame, so even if the man walks to the right and moves, it fits firmly in the frame.



Also equipped with 'VIRTUAL BACKGROUND' that can switch the background to a virtual background, which is also implemented in Zoom etc.



Even if the child behind me is playing the toy piano messed up, using 'DENOISE' makes it completely inaudible. On the other hand, the female voice is a little muffled, so there is no problem with conversation.



'CONVERSATIONAL AI AVATAR' can convert the user's face into an

avatar .



'TRANSLATION' transcribes and translates voice in real time. It is unknown at the time of writing the article whether it also supports Japanese.



By using 'AI VIDEO COMPRESSION', it is possible to compress the video and significantly reduce the amount of data. Bandwidth usage will be reduced by up to 1/10 of

H.264 .



'FACE ALIGNMENT' is a function that makes the image look as if the speaker is looking straight at the camera, even when the speaker is not looking directly at the camera.



Maxine's data processing is done in the cloud rather than locally, so users can use these features without having to prepare a high-spec PC.

'Video conferencing is now a part of everyday life, with millions of people working, learning and playing,' said Ian Buck, general manager and vice president of accelerated computing at NVIDIA. And, by extension, video conferencing for hospital visits. Maxine integrates NVIDIA's most advanced video, audio, and interactive AI technologies with unmatched efficiency and new features. It will bring us. '

Maxine is a toolkit for third-party developers and companies to embed and use in their services, rather than a platform that consumers use directly like Zoom, and NVIDIA targets AI developers and PC makers. We have started accepting applications for early access to Maxine.

in Software,   Web Service,   Video, Posted by log1l_ks