Processing

Please wait...

Settings

Settings

Goto Application

1. WO2022072940 - PROCESSING IMAGES USING SELF-ATTENTION BASED NEURAL NETWORKS

Publication Number WO/2022/072940
Publication Date 07.04.2022
International Application No. PCT/US2021/053424
International Filing Date 04.10.2021
IPC
G06N 3/04 2006.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
G06N 3/08 2006.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
CPC
G06K 9/6267
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
62Methods or arrangements for recognition using electronic means
6267Classification techniques
G06N 3/0454
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0454using a combination of multiple neural nets
G06N 3/08
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
G06N 3/084
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
084Back-propagation
G06T 2207/20081
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
2207Indexing scheme for image analysis or image enhancement
20Special algorithmic details
20081Training; Learning
G06T 2207/20084
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
2207Indexing scheme for image analysis or image enhancement
20Special algorithmic details
20084Artificial neural networks [ANN]
Applicants
  • GOOGLE LLC [US]/[US]
Inventors
  • HOULSBY, Neil Matthew Tinmouth
  • GELLY, Sylvain
  • USZKOREIT, Jakob D.
  • ZHAI, Xiaohua
  • HEIGOLD, Georg
  • BEYER, Lucas Klaus
  • KOLESNIKOV, Alexander
  • MINDERER, Matthias Johannes Lorenz
  • WEISSENBORN, Dirk
  • DEGHANI, Mostafa
  • DOSOVITSKIY, Alexey
  • UNTERTHINER, Thomas
Agents
  • PORTNOV, Michael
Priority Data
63/087,13502.10.2020US
Publication Language English (en)
Filing Language English (EN)
Designated States
Title
(EN) PROCESSING IMAGES USING SELF-ATTENTION BASED NEURAL NETWORKS
(FR) TRAITEMENT D'IMAGES À L'AIDE DE RÉSEAUX DE NEURONES BASÉ SUR L'AUTO-ATTENTION
Abstract
(EN) Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images using self-attention based neural networks. One of the methods includes obtaining one or more images comprising a plurality of pixels; determining, for each image of the one or more images, a plurality of image patches of the image, wherein each image patch comprises a different subset of the pixels of the image; processing, for each image of the one or more images, the corresponding plurality of image patches to generate an input sequence comprising a respective input element at each of a plurality of input positions, wherein a plurality of the input elements correspond to respective different image patches; and processing the input sequences using a neural network to generate a network output that characterizes the one or more images, wherein the neural network comprises one or more self-attention neural network layers.
(FR) La présente invention concerne des procédés, des systèmes et un appareil, y compris des programmes informatiques codés sur des supports de stockage informatique, pour traiter des images à l’aide de réseaux neuronaux à auto-attention. L'un des procédés consiste à obtenir une ou plusieurs images comprenant une pluralité de pixels ; à déterminer, pour chaque image de la ou des images, une pluralité de correctifs d'image de l'image, chaque correctif d'image comprenant un sous-ensemble différent des pixels de l'image ; à traiter, pour chaque image de la ou des images, la pluralité correspondante de correctifs d'image pour générer une séquence d'entrée comprenant un élément d'entrée respectif au niveau de chaque position d'entrée d'une pluralité de positions d'entrée, une pluralité des éléments d'entrée correspondant à différents correctifs d'image respectifs ; et à traiter les séquences d'entrée à l'aide d'un réseau de neurones pour générer une sortie de réseau qui caractérise la ou les images, le réseau de neurones comprenant une ou plusieurs couches de réseau de neurones à auto-attention.
Latest bibliographic data on file with the International Bureau