CPC G06V 20/56 (2022.01) [G05D 1/0248 (2013.01); G05D 1/0257 (2013.01); G06N 20/00 (2019.01)] | 20 Claims |
1. A computer-implemented method for training a first machine-learning algorithm (MLA) to detect objects in sensor data acquired by a second sensor located at a second distance from the objects, the first MLA having been trained to recognize the objects in sensor data acquired by a first sensor located at a first distance from the objects, the second distance being greater than the first distance, the method being executed at a server, the method comprising:
receiving, by the server, first sensor data having been acquired at the first distance by the first sensor mounted on a first vehicle, the first vehicle having travelled on a predetermined road;
receiving, by the server, second sensor data having been acquired at the second distance by the second sensor mounted on a second vehicle, the second vehicle having travelled behind the first vehicle on the predetermined road;
receiving, by the server, a 3D map of the predetermined road, the 3D map including a set of objects;
aligning, by the server, the first sensor data with the second sensor data based at least in part on the set of objects in the 3D map to obtain aligned first sensor data and aligned second sensor data, such that a given region in the aligned first sensor data corresponds to a given region in the aligned second sensor data;
determining, by the first MLA, respective portions in the aligned first sensor data corresponding to respective objects, the determining including determining respective classes of the respective objects;
assigning, by the server, the respective objects and the respective object classes to respective portions in the aligned second sensor data corresponding to the respective portions in the aligned first sensor data; and
training, by the server, the first MLA to determine the objects and the respective object classes in given sensor data acquired at the second distance from the objects, the training being based on:
the respective portions of the aligned second sensor data, and
the respective object classes assigned to the respective portions of the aligned second sensor data, wherein the respective object classes were determined based on the first sensor data.
|