Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020115066 - METHOD FOR ONLINE TRAINING OF AN ARTIFICIAL INTELLIGENCE (AI) SENSOR SYSTEM

Publication Number WO/2020/115066
Publication Date 11.06.2020
International Application No. PCT/EP2019/083536
International Filing Date 03.12.2019
IPC
G06N 3/04 2006.01
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 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
G06K 9/00 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR 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
CPC
G06K 9/00624
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
00624Recognising scenes, i.e. recognition of a whole field of perception; recognising scene-specific objects
G06N 3/04
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
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
Applicants
  • IEE INTERNATIONAL ELECTRONICS & ENGINEERING S.A. [LU]/[LU]
Inventors
  • BEISE, Hans-Peter
  • SCHRÖDER, Udo
  • DIAS DA CRUZ, Steve
Agents
  • BEISSEL, Jean
  • KIHN, Pierre
  • KIHN, Henri
  • LAMBERT, Romain
  • OCVIRK, Philippe
Priority Data
LU10102806.12.2018LU
Publication Language English (EN)
Filing Language English (EN)
Designated States
Title
(EN) METHOD FOR ONLINE TRAINING OF AN ARTIFICIAL INTELLIGENCE (AI) SENSOR SYSTEM
(FR) PROCÉDÉ D'APPRENTISSAGE EN LIGNE D'UN SYSTÈME DE CAPTEUR D'INTELLIGENCE ARTIFICIELLE (IA)
Abstract
(EN)
A method of operating an artificial intelligence sensor system (10) is presented for supervised training purposes, the artificial intelligence sensor system (10) having one or more sensors (12, 14) and at least one artificial neural network (18) that is configured for receiving and processing signals (xA,1, xA,2, xB,1) from the sensor or the sensors (12, 14). The at least one artificial neural network (18) derives an output representing a quality, which encompasses abstract objects such as classes as used for classification purposes as well as properties of objects, with a confidence level regarding the provided signals (xA,1, xA,2, xB,1),. If the derived confidence level of the quality is lower than a predetermined confidence level, the at least one provided signal (xA,1) and the derived quality are temporarily stored (60). By using at least one independent sensor signal (xA,1, xB,1), which includes using a signal (xB,1) of another sensor (14), the quality having a derived confidence level lower than the predetermined confidence level is confirmed (62, 66), and the temporarily stored signal (xA,1) or signals and the confirmed quality are permanently stored (70) as labeled online training data, using the derived quality as the label.
(FR)
La présente invention concerne un procédé de fonctionnement d'un système de capteur d'intelligence artificielle (10) conçu à des fins d'apprentissage supervisé, le système de capteur d'intelligence artificielle (10) ayant un ou plusieurs capteurs (12, 14) et au moins un réseau neuronal artificiel (18) qui est conçu pour recevoir et traiter des signaux (x A, 1, x A ,2, x B ,1) envoyés par le ou les capteurs (12, 14). Ledit réseau neuronal artificiel (18) dérive une sortie représentant une qualité, qui englobe des objets abstraits tels que des catégories telles qu'utilisées à des fins de classification, ainsi que des propriétés d'objets, avec un niveau de confiance concernant les signaux fournis (x A, 1, x A ,2, x B ,1). Si le niveau de confiance dérivé de la qualité est inférieur à un niveau de confiance prédéterminé, ledit signal fourni (x A, 1) et la qualité dérivée sont temporairement stockés (60). En utilisant au moins un signal de capteur indépendant (x A, 1, x B ,1), qui comprend l'utilisation d'un signal (x B ,1) d'un autre capteur (14), la qualité ayant un niveau de confiance dérivé inférieur au niveau de confiance prédéterminé est confirmée (62, 66), et le ou les signaux stockés temporairement (x A, 1) et la qualité confirmée sont stockés de façon permanente (70) comme données d'apprentissage en ligne étiquetées, en utilisant la qualité dérivée comme étiquette.
Latest bibliographic data on file with the International Bureau