US 12,406,492 B2
Method and device for training object recognizer
Seunghun Cheon, Seongnam-si (KR)
Assigned to HL Klemove Corp., Incheon (KR)
Filed by HL Klemove Corp., Incheon (KR)
Filed on Oct. 7, 2022, as Appl. No. 17/961,907.
Claims priority of application No. 10-2021-0143049 (KR), filed on Oct. 25, 2021.
Prior Publication US 2023/0125692 A1, Apr. 27, 2023
Int. Cl. G06V 10/98 (2022.01); G06V 10/766 (2022.01); G06V 20/58 (2022.01)
CPC G06V 10/98 (2022.01) [G06V 10/766 (2022.01)] 11 Claims
OG exemplary drawing
 
1. A method for training an object recognizer, the method comprising:
obtaining an image by capturing an object by a first sensor and a second sensor;
obtaining first object recognition information by inputting an image captured by the first sensor to a first sensor-based object recognizer and obtaining second object recognition information by inputting an image captured by the second sensor to a second sensor-based object recognizer;
detecting an object recognition error in the second sensor-based object recognizer;
if the object recognition error is detected, obtaining a predicted value of the second object recognition information corresponding to the first object recognition information based on reference data created before; and
training the second sensor-based object recognizer using the predicted value of the second object recognition information,
wherein the predicted value of the second object recognition information is obtained by a regression model representing a relationship between the first object recognition information and the second object recognition information,
wherein the regression model is represented by an equation in which the second object recognition information corresponds to a value obtained by multiplying the first object recognition information by a weight and adding a bias value,
wherein the weight and the bias value are updated by training the regression model to minimize a cost function defined as a residual sum of squares (RSS) for the difference between a value of the second object recognition information in the reference data and the predicted value of the second object recognition information obtained from the regression model.