US 12,299,964 B2
Method for generating training data for a trainable method
Christopher Herbon, Weil im Schoenbuch (DE)
Assigned to ROBERT BOSCH GMBH, Stuttgart (DE)
Filed by Robert Bosch GmbH, Stuttgart (DE)
Filed on Jan. 12, 2023, as Appl. No. 18/153,529.
Claims priority of application No. 10 2022 200 488.4 (DE), filed on Jan. 18, 2022; and application No. 10 2022 209 401.8 (DE), filed on Sep. 9, 2022.
Prior Publication US 2023/0230360 A1, Jul. 20, 2023
Int. Cl. G06V 20/70 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 20/58 (2022.01)
CPC G06V 10/774 (2022.01) [G06V 10/776 (2022.01); G06V 20/58 (2022.01); G06V 20/70 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A method for generating training data for a trainable method for a system which includes one or multiple sensors for detecting at least one subarea of the surroundings around the system, the method comprising the following steps:
a) obtaining one first detection and at least one second detection having at least one known relative ratio between the first and second detections and/or the sensors that have carried out the first and second detections;
b) determining at least a portion of content of each of the first and second detections, and assigning at least one piece of information concerning the determined content to each of the the first and second detections;
c) projecting the at least one assigned piece of information from one of the first and second detections and/or from a content representation associated with one of the first and second detections into at least one other of the first and second detections and/or into a content representation associated with the at least one other of the first and second detections;
d) checking at least one subarea of at least one of the first and second detections and/or of at least one of the content representations for possible inconsistencies in the determined content of the first and second detections.