US 11,670,010 B2
Data compression using conditional entropy models
David Charles Minnen, Mountain View, CA (US); Saurabh Singh, Mountain View, CA (US); Johannes Balle, San Francisco, CA (US); Troy Chinen, Newark, CA (US); Sung Jin Hwang, Mountain View, CA (US); Nicholas Johnston, San Jose, CA (US); and George Dan Toderici, Mountain View, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Jan. 19, 2022, as Appl. No. 17/578,794.
Application 17/578,794 is a continuation of application No. 16/515,586, filed on Jul. 18, 2019, granted, now 11,257,254.
Claims priority of provisional application 62/701,264, filed on Jul. 20, 2018.
Prior Publication US 2022/0138991 A1, May 5, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 9/00 (2006.01); G06N 20/00 (2019.01); G06F 17/18 (2006.01); G06N 3/08 (2023.01); G06T 3/40 (2006.01)
CPC G06T 9/001 (2013.01) [G06F 17/18 (2013.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01); G06T 3/40 (2013.01); G06T 9/002 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method implemented by one or more data processing apparatus for decompressing a compressed representation of a set of data, the method comprising:
obtaining the compressed representation of the data, wherein the compressed representation of the data comprises respective entropy encoded representations of: (i) a latent representation of the data, and (ii) a latent representation of an entropy model used to entropy encode the latent representation of the data;
entropy decoding the latent representation of the entropy model;
entropy decoding the latent representation of the data, comprising:
processing the latent representation of the entropy model using a hyper-decoder neural network to generate a reconstruction of the entropy model, wherein the reconstruction of the entropy model defines one or more code symbol probability distributions; and
entropy decoding the latent representation of the data using the code symbol probability distributions defined by the reconstruction of the entropy model; and
processing the latent representation of the data using a decoder neural network to generate a reconstruction of the data.