US 12,346,991 B2
Low- and high-fidelity classifiers applied to road-scene images
Vidya Nariyambut Murali, Sunnyvale, CA (US); and Madeline Jane Schrier, Palo Alto, CA (US)
Assigned to Ford Global Technologies, LLC, Dearborn, MI (US)
Filed by Ford Global Technologies, LLC., Dearborn, MI (US)
Filed on Jul. 11, 2023, as Appl. No. 18/350,354.
Application 18/350,354 is a continuation of application No. 17/477,282, filed on Sep. 16, 2021, granted, now 11,734,786.
Application 17/477,282 is a continuation of application No. 16/444,301, filed on Jun. 18, 2019, granted, now 11,200,447.
Application 16/444,301 is a continuation of application No. 14/995,134, filed on Jan. 13, 2016, granted, now 10,373,019.
Prior Publication US 2023/0351544 A1, Nov. 2, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 10/44 (2022.01); G06F 18/20 (2023.01); G06T 1/20 (2006.01); G06V 10/70 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/58 (2022.01); G06V 30/19 (2022.01)
CPC G06T 1/20 (2013.01) [G06F 18/285 (2023.01); G06V 10/454 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 10/87 (2022.01); G06V 20/582 (2022.01); G06V 30/19113 (2022.01); G06T 2207/30252 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
down-sampling an input image to generate a down sampled image;
extracting a plurality of regions from the down sampled image;
forward feeding each of the plurality of regions through a low-fidelity classifier of a neural network to identify one or more candidate regions depicting an object of interest; and
forward feeding the one or more candidate regions through a high-fidelity classifier of the neural network to confirm whether any of the one or more candidate regions depicts the object of interest;
wherein down-sampling the input image comprises maintaining one or more of a predetermined percent of entropy, a ratio of entropy, or an absolute value of entropy in the down sampled image.