Method and control device for training an object detector

Patent metadata

Field	Value
Publication number	US-12534104-B2
Application number	US-202318157544-A
Country	US
Kind code	B2
Filing date	Jan 20, 2023
Priority date	Jan 24, 2022
Publication date	Jan 27, 2026
Grant date	Jan 27, 2026

Abstract

A method is for training an object detector configured to detect objects in sensor data of a sensor. The method includes providing first sensor data of the sensor, providing an object representation assigned to the first sensor data, and transmitting the object representation to a sensor model. The method further includes imaging object representations onto the first sensor data of the sensor with the sensor model, assigning the object representation to second sensor data with the sensor model, and training the object detector based on the second sensor data.

First claim

What is claimed is:

1 . A method for training an object detector model configured to detect objects in sensor data of a first sensor, the method comprising:
providing first sensor data of the first sensor; providing an object representation assigned to the first sensor data, the object representation including an object detection result generated by the object detector model based on the first sensor data; transmitting the object representation to a sensor model, the sensor model describing measurement characteristics of the first sensor; determining second sensor data including values to be expected to be measured by for the first sensor by imaging the object representation onto the first sensor data of the first sensor with the sensor model; assigning the object representation to the second sensor data with the sensor model; and training the object detector model based on the second sensor data, the object detector model being a neural network having a plurality of network weights that are adapted during the training, the training the object detector model including (i) providing annotations, (ii) comparing the object detection result with the annotations in order to determine a cost function, and (iii) adapting the plurality of network weights based on the cost function.

2 . The method according to claim 1 , wherein training the object detector model further comprises:
comparing the second sensor data with the first sensor data in order to determine another cost function; and adapting the plurality of network weights based on the other cost function.

3 . The method according to claim 1 , wherein the annotations are generated using a second sensor.

4 . The method according to claim 1 , wherein the sensor model is an artificial neural network.

5 . A method for controlling a driver assistance system of a motor vehicle, comprising:
training an object detector model configured to detect objects in sensor data of a first sensor of the motor vehicle, the object detector model being a neural network having a plurality of network weights that are adapted during the training, the training including:
providing first sensor data of the first sensor,
providing an object representation assigned to the first sensor data, the object representation including an object detection result generated by the object detector model based on the first sensor data,
transmitting the object representation to a sensor model, the sensor model describing measurement characteristics of the first sensor,
determining second sensor data including values to be expected to be measured by the first sensor by imaging the object representation onto the first sensor data of the first sensor with the sensor model,
assigning the object representation to the second sensor data with the sensor model, and
training the object detector model based on the second sensor data, the training the object detector model including (i) providing annotations, (ii) comparing the object detection result with the annotations in order to determine a cost function, and (iii) adapting the plurality of network weights based on the cost function;
providing the trained object detector model for the first sensor of the motor vehicle; generating object detection results using the trained object detector model; and controlling the driver assistance system based on the object detection results.

6 . A control device for training an object detector model for detecting objects in sensor data of a first sensor, the control device comprising:
a first receiver configured to provide first sensor data of the first sensor; a second receiver configured to provide an object representation assigned to the first sensor data, the object representation including an object detection result generated by the object detector model based on the first sensor data; a transmitter configured to transmit the object representation to a sensor model configured to determine second sensor data including values to be expected to be measured by the first sensor by imaging the object representation onto the first sensor data of the first sensor, the sensor model describing measurement characteristics of the first sensor; and a processor configured to (i) assign the object representation to the second sensor data using the sensor model and (ii) train the object detector model based on the second sensor data, the object detector model being a neural network having a plurality of network weights that are adapted during the training, the training the object detector model including providing annotations, comparing the object detection result with the annotations in order to determine a cost function, and adapting the plurality of network weights based on the cost function.

7 . The control device according to claim 6 , the processor further configured to:
compare the second sensor data with the first sensor data in order to determine another cost function; and adapt the plurality of network weights based on the other cost function.

8 . The control device according to claim 6 , wherein the object representation is an annotation.

9 . The control device according to claim 6 , wherein the sensor model is an artificial neural network.

10 . The control device according to claim 6 , wherein a driver assistance system of a motor vehicle including the first sensor is controlled by (i) receiving the object detector model for the first sensor of the motor vehicle, (ii) generating object detection results using the object detector model, and (iii) controlling the driver assistance system based on the generated object detection results.

Assignees

• Bosch Gmbh Robert

Inventors

• Glaeser Claudius
• Timm Fabian
• Drews Florian
• Ulrich Michael
• Faion Florian
• Rosenbaum Lars

Classifications

• B60W2420/00

  Indexing codes relating to the type of sensors based on the principle of their operation · CPC title

• G06V20/58

  Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads · CPC title

• G06V10/82 Primary

  using neural networks · CPC title

• G06V10/751

  Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching · CPC title

• B60W50/06

  Improving the dynamic response of the control system, e.g. improving the speed of regulation or avoiding hunting or overshoot · CPC title