US 12,233,912 B2
Efficient neural networks
Nikhil Nagraj Rao, Sunnyvale, CA (US); Francois Charette, Tracy, CA (US); Shruthi Venkat, Sunnyvale, CA (US); Sandhya Sridhar, Sunnyvale, CA (US); and Vidya Nariyambut Murali, Sunnyvale, CA (US)
Assigned to Ford Global Technologies, LLC, Dearborn, MI (US)
Filed by Ford Global Technologies, LLC, Dearborn, MI (US)
Filed on Jan. 10, 2022, as Appl. No. 17/571,944.
Prior Publication US 2023/0219601 A1, Jul. 13, 2023
Int. Cl. B60W 60/00 (2020.01); G06V 10/82 (2022.01); G06V 20/56 (2022.01)
CPC B60W 60/0025 (2020.02) [G06V 10/82 (2022.01); G06V 20/56 (2022.01); B60W 2556/55 (2020.02); B60W 2756/10 (2020.02)] 20 Claims
OG exemplary drawing
 
1. A system, comprising a first computer that includes
a first processor; and
a first memory, the first memory including first instructions executable by the first processor to:
determine a location of a first object in an image;
draw a line on the image based on the location of the first object;
train a deep neural network to determine a relative location between the first object in the image and a second object in the image based on determining overshoot between the second object and the first object based on the line; and
optimize the deep neural network by determining a fitness score that divides a plurality of deep neural network parameters by a performance score.