| CPC G01S 13/867 (2013.01) [B60W 30/08 (2013.01); G01S 7/417 (2013.01); G01S 13/86 (2013.01); G01S 13/89 (2013.01); G01S 13/931 (2013.01); G06T 7/70 (2017.01); B60W 2420/403 (2013.01); B60W 2420/408 (2024.01); B60W 2554/00 (2020.02); B60W 2754/10 (2020.02); G01S 2013/93271 (2020.01); G06T 2207/10024 (2013.01); G06T 2207/10028 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30252 (2013.01)] | 20 Claims |
|
1. A method comprising:
receiving, at a computing system coupled to a vehicle, radar data representing an environment;
detecting, based on the radar data, an object positioned in the environment;
receiving, at the computing system, an image representing the environment;
based on detecting the object, removing peripheral areas of the image to generate an image patch that focuses upon the object;
determining, based on the image patch, a first set of channels comprising a red pixel channel, a green pixel channel, and a blue pixel channel;
determining, based on radar data corresponding to the object, a second set of channels comprising a heading channel and a range channel;
providing the first set of channels and the second set of channels as inputs into a neural network; and
providing, based on an output from the neural network, information about the object via one or more peripherals of the vehicle, wherein the output from the neural network includes a first channel indicating a classification for the object, a second channel indicating a center of the object from a perspective of the vehicle, and one or more channels indicating dimensions for the object from the perspective of the vehicle.
|