US 11,941,918 B2
Extracting information from images
Symeon Nikitidis, London (GB); Francisco Angel Garcia Rodriguez, London (GB); Erlend Davidson, London (GB); and Samuel Neugber, London (GB)
Assigned to Yoti Holding Limited, London (GB)
Filed by Yoti Holding Limited, London (GB)
Filed on Apr. 14, 2023, as Appl. No. 18/301,147.
Application 18/301,147 is a continuation of application No. 17/107,654, filed on Nov. 30, 2020, granted, now 11,657,525.
Application 17/107,654 is a continuation in part of application No. 17/058,371, granted, now 11,281,921, issued on Mar. 22, 2022, previously published as PCT/EP2019/083712, filed on Dec. 4, 2019.
Claims priority of application No. 1819794 (GB), filed on Dec. 4, 2018.
Prior Publication US 2023/0252662 A1, Aug. 10, 2023
Int. Cl. G06V 40/40 (2022.01); B29C 64/393 (2017.01); B33Y 50/00 (2015.01); B33Y 50/02 (2015.01); G06F 18/214 (2023.01); G06N 3/08 (2023.01); G06T 7/20 (2017.01); G06T 7/50 (2017.01); G06T 7/55 (2017.01); G06T 17/00 (2006.01); G06V 10/141 (2022.01); G06V 10/44 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/64 (2022.01); G06V 40/16 (2022.01); B29L 31/48 (2006.01); B33Y 30/00 (2015.01); B33Y 80/00 (2015.01); G06F 21/32 (2013.01)
CPC G06V 40/40 (2022.01) [B29C 64/393 (2017.08); B33Y 50/00 (2014.12); B33Y 50/02 (2014.12); G06F 18/214 (2023.01); G06N 3/08 (2013.01); G06T 7/20 (2013.01); G06T 7/50 (2017.01); G06T 7/55 (2017.01); G06T 17/00 (2013.01); G06V 10/141 (2022.01); G06V 10/454 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/647 (2022.01); G06V 40/16 (2022.01); G06V 40/168 (2022.01); G06V 40/172 (2022.01); B29L 2031/48 (2013.01); B33Y 30/00 (2014.12); B33Y 80/00 (2014.12); G06F 21/32 (2013.01); G06T 2200/04 (2013.01); G06T 2207/10021 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30201 (2013.01)] 10 Claims
OG exemplary drawing
 
1. An anti-spoofing system comprising:
a data store comprising computer-executable instructions; and
a processor configured to execute the computer-executable instructions to implement:
a depth estimation component configured to receive a 2D verification image captured by a 2D image capture device and to extract estimated depth information therefrom;
a global anti-spoofing classifier configured to use the extracted depth information to classify the 2D verification image in relation to real and spoofing classes corresponding, respectively, to 2D verification images of actual humans and 2D verification images of spoofing entities, and thereby assign a global classification value to the whole of the 2D verification image; and
a patch-based anti-spoofing classifier configured to classify each image patch of multiple image patches of the 2D verification image in relation to the real and spoofing classes, and thereby assign a local classification value to each image patch of the multiple image patches;
wherein the anti-spoofing system is configured to use the global and local classification values to determine whether an entity captured in the 2D verification image corresponds to an actual human or a spoofing entity;
wherein the patch-based anti-spoofing classifier has a convolutional neural network (CNN) architecture, wherein each of the multiple image patches is defined by a configuration of convolutional filtering layers within the CNN architecture;
wherein the configuration of the convolutional filtering layers is such that the image patches are overlapping.