US 11,720,622 B2
Machine learning multiple features of depicted item
Oren Barkan, Bishon Lezion (IL); Noam Razin, Jerusalem (IL); Noam Koenigstein, Tel Aviv (IL); Roy Hirsch, Ramat Yishai (IL); and Nir Nice, Salit (IL)
Assigned to Microsoft Technology Licensing, LLC
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jun. 9, 2022, as Appl. No. 17/836,779.
Application 17/836,779 is a continuation of application No. 16/725,652, filed on Dec. 23, 2019, granted, now 11,373,095.
Prior Publication US 2022/0300814 A1, Sep. 22, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 16/532 (2019.01); G06N 3/08 (2023.01); G06N 20/20 (2019.01); G06N 20/10 (2019.01); G06T 7/00 (2017.01); G06F 17/18 (2006.01); G06N 3/045 (2023.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01)
CPC G06F 16/532 (2019.01) [G06F 17/18 (2013.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06N 20/10 (2019.01); G06N 20/20 (2019.01); G06T 7/0002 (2013.01); G06T 7/97 (2017.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06T 2207/20076 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 21 Claims
OG exemplary drawing
 
1. A computing system that trains a neural network to identify machine recognizable features of an item that is embodied in an image and to use the neural network to identify other items that have similar features to the machine recognizable features, said computing system comprising:
at least one processor; and
at least one hardware storage device that stores instructions that are executable by the at least one processor to cause the computing system to:
access a plurality of images, wherein each image in the plurality of images provides a different visualization of a same item;
machine train on the plurality of images using a neural network to identify a plurality of features of the item;
generate a plurality of embedding vectors for each feature in the plurality of features such that the neural network is trained on multiple features of the item, wherein the plurality of embedding vectors includes an identity embedding vector that provides a supposed identity for the item; and
use the identity embedding vector to generate a probability vector representing probabilities that the supposed identity of the item is of various values.