US 12,327,417 B1
Drowsy driving detection
Sung Chun Lee, Fairfax, VA (US); Nathan Hurst, Seattle, WA (US); Yan Wang, Mercer Island, WA (US); Olamide Akintewe, Potterville, MN (US); Justin Levine, Erie, CO (US); Kenshiro Nakagawa, San Francisco, CA (US); Cole Jurden, Kansas City, MO (US); Rachel Demerly, New York, NY (US); Aravindh Ramesh, San Francisco, CA (US); Kevin Lai, Redmond, WA (US); Jovanna Bubar, Los Angeles, CA (US); Shirish Nair, Shoreline, WA (US); and Maisie Wang, Seattle, WA (US)
Assigned to Samsara Inc., San Francisco, CA (US)
Filed by Samsara Inc., San Francisco, CA (US)
Filed on Aug. 19, 2024, as Appl. No. 18/808,818.
Application 18/808,818 is a continuation of application No. 18/634,353, filed on Apr. 12, 2024, granted, now 12,112,555.
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 20/59 (2022.01); G06V 10/774 (2022.01); G06V 40/16 (2022.01)
CPC G06V 20/597 (2022.01) [G06V 10/774 (2022.01); G06V 40/161 (2022.01); G06V 40/168 (2022.01); G06V 40/174 (2022.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
training a drowsiness model with data associated with inward videos of cabins of vehicles captured during trips of the vehicles;
obtaining a plurality of inward frames during a trip of a first vehicle;
calculating a plurality of embeddings for the plurality of inward frames, each embedding being a transformation of the inward frame into a vector;
providing the plurality of embeddings of the plurality of inward frames as an input to the drowsiness model executing in a device at the first vehicle; and
calculating, by the drowsiness model, a drowsiness value based on the plurality of embeddings of the plurality of inward frames, the drowsiness value indicating a level of drowsiness of a driver of the first vehicle.