There is provided a computer-implemented method of generating training data for training a machine learning model to detect automatically a lane boundary from an image of a route traversed by an autonomous vehicle (Fig. 12, S402). Obtaining an image captured by an autonomous vehicle during a traversal of the route S300; projecting a lane boundary model, from a three-dimensional LiDAR point cloud of the route S302, on to the image and generating a training data example for training the machine learning model by removing automatically one or more sections of the lane boundary corresponding to one or more respective occlusions in the image S304. The machine learning model may be a convolutional neural network and the distance between cloud points may be clustered based on a distance threshold.


    Access

    Download


    Export, share and cite



    Title :

    A computer-implemented method of generating training data for training a machine learning model


    Publication date :

    2022-09-07


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    English



    Method, apparatus and computer program for training machine learning model and generating training data

    KNOBLOCH DANIEL / SCHULZ BRITTA | European Patent Office | 2022

    Free access

    Automatically generating machine-learning training data

    PARENTI ROBERT / SIDDIQUI ADIL NIZAM / NEUBECKER CYNTHIA M | European Patent Office | 2025

    Free access

    AUTOMATICALLY GENERATING MACHINE-LEARNING TRAINING DATA

    PARENTI ROBERT / SIDDIQUI ADIL NIZAM / NEUBECKER CYNTHIA M | European Patent Office | 2023

    Free access

    COMPUTER-IMPLEMENTED METHOD FOR TRAINING OF A MACHINE LEARNING MODEL IN VEHICLE ACCIDENT EVENT ASSESSMENT

    ALKHOURY ZIAD / AHMED JAWWAD | European Patent Office | 2024

    Free access

    COMPUTER-IMPLEMENTED METHOD FOR TRAINING OF A MACHINE LEARNING MODEL IN VEHICLE ACCIDENT EVENT ASSESSMENT

    AHMED JAWWAD / ALKHOURY ZIAD | European Patent Office | 2024

    Free access