Patent · US Active

Method for building a computer-implemented tool for assessment of qualitative features from face images

US10956716B2 · kind B2 · utility

0Cited by
1References
11Claims
0Family size

Assignee

Inventors

Key dates

Filing dateFeb 25, 2019
Grant dateMar 23, 2021
Priority date
Expiry dateMay 11, 2039

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06N3/08
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

Disclosed is a method for building a neural network tool to determine, by processing a picture of a human face, a score representative of a qualitative feature of the face, including performing a training session of an initial neural network to assess, from the picture, an estimated age of the person. The initial neural network includes layers of neurons and associated weights, and each training session is performed on a first database of face images with associated ages, and each training session outputs an updated version of the neural network with adjusted weights. For each updated version of the initial neural network corresponding to a training session, evaluating an error of part of the initial neural network on another database of images with each associated with a reference score representing the qualitative feature, selecting the part of the initial neural network in the updated version corresponding to the minimum.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.