US 12,423,973 B2
Device and computer implemented method for explainable scene clustering
Cory Henson, Pittsburgh, PA (US); Daria Stepanova, Leonberg (DE); Mohamed Gad-Elrab, Saarbruecken (DE); Ruwan Wickramarachchige Don, Columbia, SC (US); and Sreyasi Nag Chowdhury, Saarbruecken (DE)
Assigned to ROBERT BOSCH GMBH, Stuttgart (DE)
Filed by Robert Bosch GmbH, Stuttgart (DE)
Filed on Sep. 7, 2022, as Appl. No. 17/939,517.
Claims priority of application No. 10 2021 210 848.2 (DE), filed on Sep. 28, 2021.
Prior Publication US 2023/0097860 A1, Mar. 30, 2023
Int. Cl. G06K 9/00 (2022.01); G06N 5/022 (2023.01); G06V 20/40 (2022.01); G06V 20/58 (2022.01)
CPC G06V 20/41 (2022.01) [G06N 5/022 (2013.01); G06V 20/58 (2022.01)] 21 Claims
OG exemplary drawing
 
1. A computer implemented method for explainable clustering of a scene, the method comprising the following steps:
identifying objects in the scene, the objects including a first object that belongs to a first object class and a second object that belongs to a second object class;
based on the identification of the objects, identifying, in a first knowledge graph, a pair of entities that respectively correspond to the first and second object classes, wherein the first knowledge graph is a scene-independent commonsense knowledge graph that identifies a first relation, the first relation being a scene-independent relationship of how the pair of entities is related to each other without regard to the scene;
clustering the scene into one of a plurality of scene clusters, the scene cluster into which the scene is clustered also including other scenes, wherein the clustering is performed based on commonalities between the scene and the other scenes determined at least partly using the first relation;
based on the scene cluster into which the scene has been clustered, determining a second relation that relates one or more visual features of digital image data representing the scene to the scene cluster into which the scene has been clustered; and
determining a rule that maps the first relation to the second relation.