Here Global B.v.

Ian Endres

Victor Lu

Ole Henry Dorum

David Doria

Jeffrey Minoru Adachi

Antonio Haro

Shubhabrata Roy

Guangli Wu

Bin Chen

Hank Sutton

An apparatus accesses label data and training images corresponding to a geographic area; and provides the label data and training images to a training model. The training model comprises of at least a predictor model and an alignment model. The predictor model is configured to receive an image and provide a prediction corresponding to the image. The alignment model is configured to generate a transformed prediction based on aligning the label data and the prediction. The apparatus executes a loss engine to iteratively receive the label data and the transformed prediction, evaluate a loss function based on the label data and the transformed prediction, and cause weights of the predictor model and the alignment model to be updated based on the evaluated loss function to cause the predictor and alignment models to be trained.

Semantic segmentation ground truth correction with spatial transformer networks

Segmentation of three dimensional objects may be implemented using a neural network model, a clustering module, a factorization module, and a geometric fitting module. The neural network model is configured to analyze point cloud data for a geographic region and assign probability values outputted from the neural network to points in the point cloud data. The clustering module is configured to group a subset of the probability values based on relative locations of the assigned points in the point cloud data. The factorization module is configured to factor a matrix with the subset of the clustered probability values to assign a line for a three dimensional object of the geographic region. The geometric fitting module is configured to fit at least one predetermined shape for the three dimensional object to the point cloud data based at least on the assigned line.

Scalable three dimensional object segmentation

A geographic database storing map data is provided. The geographic database is stored in a non-transitory computer readable medium. The geographic database comprises a plurality of records corresponding to drivable surfaces of a road network. The plurality of records comprise a plurality of lane records corresponding to particular lanes of the road network. Each first record of the plurality of records comprises a plurality of instances of adjacency information. Each instance of adjacency information/data (a) links the first record corresponding to a first drivable surface of the road network to a second record of the plurality of records corresponding to a second drivable surface of the road network. The first drivable surface is adjacent to the second drivable surface. Each instance of adjacency information/data indicates crossing parameters between the first drivable surface and the second drivable surface.

Lane-centric road network model for navigation

Embodiments described herein relate generally to determining correspondence between a template and an object in an image. A method may include: receiving an image of an environment including an image of an object within the image of the environment; resizing the first template to obtain a scaled first template having a size corresponding to a size of the image of the object; calculating a number of correspondences between the scaled first template and the image of the object; receiving a candidate homography; testing the candidate homography; and replacing the image of the object with a second template of a different object according to the candidate homography in response to the candidate homography being established as corresponding to the image of the object.

Method and apparatus for homography estimation

An approach is provided for generating a cleaned object model to represent an object in a mapping database. The approach includes receiving point cloud data depicting the object. The approach also includes processing the point cloud data to determine one or more surface points of the point cloud data. The one or more surface points represent one or more surfaces of the object. The approach further includes cutting a model of the object into one or more fragments using the one or more surface points. The one or more fragments include one or more object fragments and one or more non-object fragments. The approach further includes designating the one or more object fragments as the cleaned object model to represent the object in the mapping database.

Method and apparatus for generating a cleaned object model for an object in a mapping database

A method, apparatus and computer program product are provided for generating map geometry based on a received image and probe data. A method is provided including receiving a first image and probe data associated with the first image, categorizing pixels of the first image based on the probe data, and generating a map geometry based on the pixel categorization of the first image.

Growing community of inventors

Naperville, IL, United States of America

Ian Endres