Methods and Systems for Training a Machine-Learning Method

    公开(公告)号:US20220114489A1

    公开(公告)日:2022-04-14

    申请号:US17495332

    申请日:2021-10-06

    IPC分类号: G06N20/00

    摘要: A computer-implemented method for training a machine-learning method comprises the following steps carried out by computer hardware components: determining measurement data from a first sensor; determining approximations of ground truths based on a second sensor; and training the machine-learning method based on the measurement data and the approximations of ground truths; wherein approximations of ground truths of lower-approximation quality have a lower effect on the training than approximations of ground truths of higher-approximation quality.

    Lane marker recognition
    5.
    发明授权

    公开(公告)号:US10997434B2

    公开(公告)日:2021-05-04

    申请号:US16372817

    申请日:2019-04-02

    IPC分类号: G06K9/00 G06T7/13 G06K9/62

    摘要: A method for lane marker recognition includes: providing a filter bank with a plurality of different pairs of filters adapted to detect the left edge and the right edge of a specific type of a lane marker, respectively; receiving an image of a road; dividing the image of a road into a plurality of image segments, wherein each image segment includes at least one row of pixels of the image of a road; and for each of the image segments: applying a plurality of the pairs of filters of the filter bank to the image segment to generate a plurality of filter outputs; and determining which of the filter outputs correspond to a lane marker by using geo-metric information and appearance based information, wherein the geometric in-formation describes allowable dimensions of a determined lane marker, and the appearance based information describes allowable pixel values of a determined lane marker.