US 12,148,220 B2
Method for providing a neural network for directly validating an environment map in a vehicle by means of sensor data
Felix Drost, Puchheim (DE); and Sebastian Schneider, Feldkirchen (DE)
Assigned to Bayerische Motoren Werke Aktiengesellschaft, Munich (DE)
Appl. No. 17/766,847
Filed by Bayerische Motoren Werke Aktiengesellschaft, Munich (DE)
PCT Filed Sep. 1, 2020, PCT No. PCT/EP2020/074300
§ 371(c)(1), (2) Date Apr. 6, 2022,
PCT Pub. No. WO2021/069146, PCT Pub. Date Apr. 15, 2021.
Claims priority of application No. 10 2019 126 874.5 (DE), filed on Oct. 7, 2019.
Prior Publication US 2024/0062550 A1, Feb. 22, 2024
Int. Cl. G06V 20/56 (2022.01); G06V 10/82 (2022.01)
CPC G06V 20/56 (2022.01) [G06V 10/82 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A method for providing a neural network for directly validating an environment map in a vehicle using sensor data, comprising:
providing valid surroundings data in a feature representation comprising map data and the sensor data, wherein the map data comprises the environment map;
providing invalid surroundings data in a feature representation comprising the map data and the sensor data; and
training a neural network with the valid surroundings data and the invalid surroundings data;
wherein the providing of the valid surroundings data and the invalid surroundings data is carried out using a Siamese neural network and comprises:
generating the feature representation using a first convolutional neural network and a second convolutional neural network, wherein:
the first convolutional neural network uses the map data comprising the environment map as input, and the second convolutional neural network uses the sensor data as input; and
determining a validity of the map data in the feature representation.