Patent · US Active

Feature extraction and machine learning for automated metadata analysis

US10878296B2 · kind B2 · utility

3Cited by
0References
23Claims
0Family size

Assignee

Inventor

Key dates

Filing dateApr 12, 2019
Grant dateDec 29, 2020
Priority date
Expiry dateJun 26, 2039

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06V2201/10
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

The invention relates to image processing systems, methods, and storage media for recognizing a scene type. The invention performs dynamic content analysis to extract features from an image and creates labels that include a text-based description of the items and the environment of the image. The invention then trains multiple predictive models and determines characterization labels for the image or scene. The invention can create multi-label classifications as well as multi-class classifiers. The text-based labels created by the invention extend generic classification labels into a domain-specific manner of defining and naming groups of images. The characterizing labels can be tagged to the image as metadata for further organization and consumption of the images or scenes.

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