29-10-2020 дата публикации
Номер: US20200341466A1
Автор:
Trung Pham,
Hang Dou,
Berta Rodriguez Hervas,
Minwoo Park,
Neda Cvijetic,
David Nister,
PHAM TRUNG,
DOU HANG,
RODRIGUEZ HERVAS BERTA,
PARK MINWOO,
CVIJETIC NEDA,
NISTER DAVID,
Pham, Trung,
Dou, Hang,
Rodriguez Hervas, Berta,
Park, Minwoo,
Cvijetic, Neda,
Nister, David
Принадлежит:
In various examples, live perception from sensors of a vehicle may be leveraged to generate potential paths for the vehicle to navigate an intersection in real-time or near real-time. For example, a deep neural network (DNN) may be trained to compute various outputs—such as heat maps corresponding to key points associated with the intersection, vector fields corresponding to directionality, heading, and offsets with respect to lanes, intensity maps corresponding to widths of lanes, and/or classifications corresponding to line segments of the intersection. The outputs may be decoded and/or otherwise post-processed to reconstruct an intersection—or key points corresponding thereto—and to determine proposed or potential paths for navigating the vehicle through the intersection. 1. A method comprising:applying, to a neural network, image data representative of an intersection in a field of view of an image sensor of a vehicle;computing, using the neural network and based at least in part on the image data, first data representative of one or more two dimensional (2D) heat maps representing locations of key points in corresponding to the intersection and second data representative of classification confidence values corresponding to the key points;decoding the one or more 2D heat maps to determine pixel locations of the key points in the image;determining associated classifications from the classification confidence values that correspond to the key points;based at least in part on the pixel locations and the associated classifications, generating one or more proposed paths through the intersection; andperforming one or more operations by the vehicle using the one or more proposed paths.2. The method of claim 1 , wherein the associated classifications represent one or more of: an entrance point to a pedestrian crossing claim 1 , an entrance point to the intersection claim 1 , an exit point from the intersection claim 1 , an exit point from a pedestrian crossing claim 1 , ...
Подробнее
