US 12,032,656 B2
Concept for generating training data and training a machine-learning model for use in re-identification
Thomas Jakobsen, Copenhagen K (DK); David Guldbrandsen, Copenhagen K (DK); Max La Cour Christensen, Copenhagen K (DK); and Ulrik Ishoej Soendergaard, Copenhagen K (DK)
Assigned to Grazper Technologies ApS, Copenhagen (DK)
Filed by Grazper Technologies ApS, Copenhagen K (DK)
Filed on Aug. 27, 2021, as Appl. No. 17/446,174.
Claims priority of application No. 20197545 (EP), filed on Sep. 22, 2020.
Prior Publication US 2022/0092348 A1, Mar. 24, 2022
Int. Cl. G06F 18/214 (2023.01); G06T 7/70 (2017.01); G06V 20/30 (2022.01); G06V 40/16 (2022.01)
CPC G06F 18/214 (2023.01) [G06T 7/70 (2017.01); G06V 20/30 (2022.01); G06V 40/172 (2022.01)] 13 Claims
OG exemplary drawing
 
1. A computer system for generating training data for training a machine-learning model for use in re-identification, the computer system comprising processing circuitry configured to:
obtain media data, the media data comprising a plurality of samples representing a person, an animal or an object,
wherein each sample is associated with secondary information characterizing the sample or the person, animal or object represented by the sample, and
wherein, if the media data comprises media data of two or more cameras, the secondary information characterizes an origin of the respective sample with respect to the two or more cameras and a capture time of the sample;
process the media data to identify tuples of samples that represent the same person, animal or object, wherein the processing circuitry is configured to identify the tuples of samples that represent the same person, animal or object based on:
the secondary information being associated with the respective samples,
the origin of the samples with respect to the two or more cameras, and
the capture time of the samples; and
generate the training data based on the identified tuples of samples that represent the same person, animal or object.