US 11,954,919 B2
Traffic light active learning pipeline
Kun-Hsin Chen, Mountain View, CA (US); Peiyan Gong, Ann Arbor, MI (US); Shunsho Kaku, Mountain View, CA (US); Sudeep Pillai, Santa Clara, CA (US); Hai Jin, Los Altos, CA (US); Sarah Yoo, Mountain View, CA (US); David L. Garber, Palo Alto, CA (US); and Ryan W. Wolcott, Ann Arbor, MI (US)
Assigned to TOYOTA RESEARCH INSTITUTE, INC., Los Altos, CA (US)
Filed by TOYOTA RESEARCH INSTITUTE, INC., Los Altos, CA (US)
Filed on Oct. 8, 2020, as Appl. No. 17/066,351.
Prior Publication US 2022/0114375 A1, Apr. 14, 2022
Int. Cl. G06V 20/58 (2022.01); G01C 21/32 (2006.01); G06N 3/08 (2023.01); H04W 4/44 (2018.01)
CPC G06V 20/584 (2022.01) [G01C 21/32 (2013.01); G06N 3/08 (2013.01); H04W 4/44 (2018.02)] 14 Claims
OG exemplary drawing
 
1. A method, comprising:
predicting a state of a traffic signal using a model;
obtaining vehicle-to-infrastructure (V2I)-based information regarding the state of the traffic signal;
comparing the predicted state of the traffic signal with the V2I-based information regarding the state of the traffic signal;
in response to a determination that the predicted state of the traffic signal is inconsistent with that of the V2I-based information regarding the state of the traffic signal, updating training data by saving data related to the state of the traffic signal; and
training the model used to predict the state of the traffic signal using the updated training data.