Image-based ontology refinement using clusters

1. A method comprising:

identifying, by performing image processing on an image, a scene and a set of objects within the scene of the image;

storing, by a processor, the set of objects in a cluster labeled by the scene;

refining, by the processor, using the cluster, a knowledge graph; and

determining a membership strength between an object of the set of objects and the cluster labeled by the scene, wherein the membership strength is determined by dividing a number of times the object was observed within the scene by a total number of times the scene was observed.

2. The method of claim 1 , further comprising:

receiving a query related to the cluster labeled by the scene;

determining whether the membership strength of the object satisfies a membership strength threshold; and

outputting, in response to a determination that the membership strength of the object satisfies the membership strength threshold, the object as a response to the query.

3. The method of claim 1 , wherein refining the knowledge graph includes:

creating a classification corresponding to the cluster within the knowledge graph;

determining whether the membership strength between the object and the cluster satisfies a membership strength threshold; and

creating, in response to determining that the membership strength between the object and the cluster satisfies the membership strength threshold, a relationship between the object and the classification within the knowledge graph.

4. The method of claim 1 , wherein refining the knowledge graph includes:

creating a classification corresponding to the cluster; and

creating a relationship between the classification and the set of objects within the knowledge graph.

5. The method of claim 1 , wherein prior to storing the set of objects in the cluster labeled by the scene, the method further comprises:

determining whether the cluster already exists for the scene;

referencing, in response to determining that the cluster already exists for the scene, the existing cluster;

determining whether an object within the set of objects is already within the cluster; and

refining, in response to determining that the object is already within the cluster, a membership strength between the object and the cluster.

6. A system comprising:

a memory storing program instructions; and

a processor, wherein the processor is configured to execute the program instructions to perform a method comprising:

identifying, using an image processing technique, a scene and a set of objects within the scene of an image;

storing the set of objects in a cluster labeled by the scene;

refining a knowledge graph using the cluster; and

determining a membership strength between an object of the set of objects and the cluster labeled by the scene, wherein the membership strength is determined by dividing a number of times the object was observed within the scene by a total number of times the scene was observed.

7. The system of claim 6 , wherein the method performed by the processor further comprises:

receiving a query related to the cluster labeled by the scene;

determining whether the membership strength of the object satisfies a membership strength threshold; and

outputting, in response to a determination that the membership strength of the object satisfies the membership strength threshold, the object as a response to the query.

8. The system of claim 6 , wherein refining the knowledge graph includes:

creating a classification corresponding to the cluster within the knowledge graph;

determining whether the membership strength between the object and the cluster satisfies a membership strength threshold; and

adding, in response to determining that the membership strength between the object and the cluster satisfies the membership strength threshold, a relationship between the object and the classification within the knowledge graph.

9. The system of claim 6 , wherein refining the knowledge graph includes:

creating a classification corresponding to the cluster; and

creating a relation between the classification and the set of objects within the knowledge graph.

10. The system of claim 6 , wherein prior to storing the set of objects in the cluster labeled by the scene, the method performed by the processor comprises:

determining a classification label confidence of each object of the set of objects; and

comparing the classification label confidence of each object of the set of objects to a classification label confidence threshold to determine a subset of objects of the set of objects which satisfy the classification label confidence threshold, wherein only the subset of objects is stored in the cluster labeled by the scene.

11. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:

identifying, by performing image processing on an image, a scene and a set of objects within the scene of the image;

storing, by a processor, the set of objects in a cluster labeled by the scene;

refining, by the processor, using the cluster, a knowledge graph; and

determining a membership strength between an object of the set of objects and the cluster labeled by the scene, wherein the membership strength is determined by dividing a number of times the object was observed within the scene by a total number of times the scene was observed.

12. The computer program product of claim 11 , wherein the method performed by the processor further comprises:

receiving a query related to the cluster labeled by the scene;

determining whether the membership strength of the object satisfies a membership strength threshold; and

outputting, in response to a determination that the membership strength of the object satisfies the membership strength threshold, the object as a response to the query.

13. The computer program product of claim 11 , wherein refining the knowledge graph includes:

creating a classification corresponding to the cluster within the knowledge graph;

determining whether the membership strength between the object and the cluster satisfies a membership strength threshold; and

creating, in response to determining that the membership strength between the object and the cluster satisfies the membership strength threshold, a relationship between the object and the classification within the knowledge graph.

14. The computer program product of claim 11 , wherein prior to storing the set of objects in the cluster labeled by the scene, the method performed by the processor comprises:

determining a classification label confidence of each object of the set of objects; and

comparing the classification label confidence of each object of the set of objects to a classification label confidence threshold to determine a subset of objects of the set of objects which satisfy the classification label confidence threshold, wherein only the subset of objects is stored in the cluster labeled by the scene.

15. The computer program product of claim 11 , wherein the program instructions are downloaded to the computer readable storage medium from a distributed data processing system.