US 11,704,563 B2
Classifying time series image data
Gaurav Kumar Singh, Westland, MI (US); Pavithra Madhavan, Westland, MI (US); Bruno Jales Costa, Sunnyvale, CA (US); Gintaras Vincent Puskorius, Novi, MI (US); and Dimitar Petrov Filev, Novi, MI (US)
Assigned to FORD GLOBAL TECHNOLOGIES, LLC, Dearborn, MI (US)
Filed by Ford Global Technologies, LLC, Dearborn, MI (US)
Filed on Apr. 27, 2021, as Appl. No. 17/241,513.
Application 17/241,513 is a continuation of application No. 16/108,698, filed on Aug. 22, 2018, granted, now 11,017,296.
Prior Publication US 2021/0248468 A1, Aug. 12, 2021
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 3/08 (2023.01); G06T 7/246 (2017.01); G06V 40/20 (2022.01); G06N 3/045 (2023.01); G06V 10/764 (2022.01); G06V 10/80 (2022.01)
CPC G06N 3/08 (2013.01) [G06N 3/045 (2023.01); G06T 7/246 (2017.01); G06V 10/764 (2022.01); G06V 10/806 (2022.01); G06V 10/809 (2022.01); G06V 40/20 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
inputting a plurality of frames of a video stream to a first neural network;
accessing first output from the first neural network;
generating one or more eccentricity maps using one or more RGB channel intensity vectors associated with the frames, wherein each eccentricity map is a static image that aggregates apparent motion of one or more features included in the video stream;
inputting the eccentricity maps to a second neural network;
accessing second output from the second neural network;
fusing the first output and second output into a fused output; and
classifying an action occurring in the video stream from the fused output.