US 12,272,171 B2
Systems and methods for image feature recognition using a lensless camera
Zhu Li, Overland Park, KS (US)
Assigned to Adeia Guides Inc., San Jose, CA (US)
Filed by Adeia Guides, Inc., San Jose, CA (US)
Filed on Sep. 21, 2023, as Appl. No. 18/371,186.
Application 18/371,186 is a continuation of application No. 17/712,325, filed on Apr. 4, 2022, granted, now 11,804,068.
Prior Publication US 2024/0135746 A1, Apr. 25, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 40/16 (2022.01); A61B 5/00 (2006.01); A61B 5/1171 (2016.01); G06F 21/32 (2013.01); G06V 10/42 (2022.01); G06V 10/77 (2022.01)
CPC G06V 40/164 (2022.01) [A61B 5/1176 (2013.01); A61B 5/7253 (2013.01); G06F 21/32 (2013.01); G06V 10/431 (2022.01); G06V 10/7715 (2022.01); G06V 40/166 (2022.01); G06V 40/169 (2022.01); G06V 40/172 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
capturing, based on light that travels to a lensless camera through apertures in a mask that comprises a pattern for masking the lensless camera, a plurality of distorted images of a user's face from different angles, wherein undistorted versions of the plurality of distorted images can only be reconstructed by a computing device with access to data defining the pattern of the mask;
using the plurality of distorted images of the user's face as a training set for training a machine learning model, where the machine learning model is configured to:
accept an input of a distorted image; and
based on the input of the distorted image, output a probability that an undistorted version of the distorted image includes a depiction of the user's face; and
inputting, by a computing device without access to data defining the pattern of the mask, a particular distorted image that is not within the training set into the machine learning model to receive as output a particular probability that an undistorted version of the particular distorted image includes a depiction of the user's face; and
performing an action based on the particular probability that the undistorted version of the particular distorted image includes a depiction of the user's face.