US 12,443,651 B2
Visual object graphs
Eric Kim, San Francisco, CA (US); and Dmitry Olegovich Kislyuk, San Ramon, CA (US)
Assigned to Pinterest, Inc., San Francisco, CA (US)
Filed by Pinterest, Inc., San Francisco, CA (US)
Filed on Jun. 29, 2023, as Appl. No. 18/344,397.
Application 18/344,397 is a division of application No. 17/848,040, filed on Jun. 23, 2022, granted, now 11,727,049.
Application 17/848,040 is a division of application No. 16/883,286, filed on May 26, 2020, granted, now 11,373,403, issued on Jun. 28, 2022.
Prior Publication US 2023/0342390 A1, Oct. 26, 2023
Int. Cl. G06F 16/55 (2019.01); G06F 17/16 (2006.01); G06F 18/22 (2023.01); G06F 18/23 (2023.01); G06T 7/10 (2017.01); G06V 20/30 (2022.01)
CPC G06F 16/55 (2019.01) [G06F 17/16 (2013.01); G06F 18/22 (2023.01); G06F 18/23 (2023.01); G06T 7/10 (2017.01); G06V 20/30 (2022.01)] 18 Claims
OG exemplary drawing
 
1. A computing system, comprising:
one or more processors; and
a memory storing program instructions that, when executed by the one or more processors, cause the one or more processors to at least:
generate, for each object of a first plurality of objects represented in a first plurality of scenes, a respective embedding vector representative of each object;
determine, based at least in part on distances between the respective embedding vectors, a first object cluster that includes a first plurality of embedding vectors that includes at least some of the respective embedding vectors and corresponds to a second plurality of objects of the first plurality of objects;
generate an object cluster embedding vector representative of the first object cluster;
generate links between the first object cluster and a second plurality of scenes of the first plurality of scenes that include one or more of the second plurality of objects, the generating links comprising determining a similarity score between each scene of the second plurality of scenes and the object cluster, the similarity score for a first scene of the second plurality of scenes being based on a similarity between an embedding vector for the object represented in the first scene and the object cluster embedding vector;
receive, from a client device, a user selection of a query object from a scene presented on the client device;
determine that the query object is associated with the first object cluster;
determine one or more first scenes from the second plurality of scenes based on the links between the first object cluster and the second plurality of scenes; and
provide at least one first scene of the one or more first scenes to the client device for presentation.