[theora] New advanced machine learning and neural networks codecs

Raphael Canut nhwcodec at gmail.com
Sun May 21 15:46:06 UTC 2017


There are more and more advanced codecs based on machine learning and
neural networks or GANs (generative adversarial networks) appearing.I have
seen some Google work and also recently WaveOne that has just unveiled some
of his work in image compression: http://www.wave.one/icml2017 .It sounds
ultra impressive! It announces compression ratios x1.7 smaller than BPG
(HEVC intra) at all quality levels for the same MS-SSIM index!!!

What is the opinion of the experts here? Is it the future of compression?
Will this affect AOM AV1 (and HEVC) development? Will the next standard
incorporate machine learning and/or neural networks? Or is it too late?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xiph.org/pipermail/theora/attachments/20170521/d6fb7840/attachment.html>

More information about the theora mailing list