Patent · US Active

Object recognition and detection using reinforcement learning

US11741152B2 · kind B2 · utility

0Cited by
7References
20Claims
0Family size

Assignee

Inventors

Key dates

Filing dateSep 30, 2020
Grant dateAug 29, 2023
Priority date
Expiry dateJun 26, 2041

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06V40/10
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

Discussed generally are techniques for managing operation of programs in a sequential order. A method can include receiving a query for an image, the query indicating characteristics of the image, selecting a chain of algorithms configured to identify the image based on the characteristics, operating an algorithm of the selected chain of algorithms that operate in increased fidelity order on an input to produce a first result, operating a ground truth algorithm on the input to generate a second result, comparing the first and second results to determine a probability of correctness (Pc) and confidence interval (CI) for the algorithm, and altering the chain of algorithms based on the determined Pc and CI.

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