US 12,322,182 B1
Visual wake content for image processing
Serdar Kizilgul, Fremont, CA (US); Steven L. Teig, Menlo Park, CA (US); and Murali Dharan, Los Altos, CA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Nov. 22, 2021, as Appl. No. 17/532,903.
Claims priority of provisional application 63/117,375, filed on Nov. 23, 2020.
Int. Cl. G06V 20/52 (2022.01); G06N 3/04 (2023.01); G06V 10/82 (2022.01); G06V 20/64 (2022.01)
CPC G06V 20/52 (2022.01) [G06N 3/04 (2013.01); G06V 10/82 (2022.01); G06V 20/64 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method of performing analysis on an image, the method comprising:
capturing a plurality of images of a scene;
for each captured image of the plurality of images:
propagating the captured image through a first plurality of layers of a first machine trained (MT) network to identify whether a predefined unique pattern is newly present in the captured image; and
when a first portion of the captured image includes a newly present predefined unique pattern:
propagating the captured image through a second plurality of layers of a second MT network to determine whether an object is within a second portion of the captured image that is identified based on the first portion that includes the newly present predefined unique pattern;
wherein, when no predefined unique pattern is identified as newly present within a particular captured image, the particular captured image is not propagated through the second plurality of layers of the second MT network, wherein the second MT network is configured to bypass the second plurality of layers when no predefined unique pattern is identified as newly present within the first portion of the captured image; and
upon determining that an object is within the second portion of the captured image, providing a notification regarding detection of the object in the image.