Method for unsupervised clustering of multimedia data using a large-scale matching system
AssigneeCortica, Ltd.
InventorsIgal Raichelgauz, Karina Ordinaev, Yehoshua Y. Zeevi
Life of this patent
Estimating the term…
Abstract
A method and apparatus for unsupervised clustering of a large-scale collection of multimedia data elements. The method comprises generating a first cluster from the large-scale collection by: matching each of the multimedia data elements to all other multimedia data elements in the large-scale collection, determining a clustering score for each match being performed, clustering multimedia data elements having a clustering score above a threshold to create the first cluster; and storing the first cluster in a storage unit.
Claims22
A method for unsupervised clustering of a large-scale collection of multimedia data elements, comprising:
- generating a first cluster from the large-scale collection by:
- matching each of the multimedia data elements to all other multimedia data elements in the large-scale collection,
- determining a clustering score for each match being performed,
- clustering multimedia data elements having a clustering score above a threshold to create the first cluster;
- wherein the generation of the first cluster includes generating for each of the multimedia data elements at least one respective signature, and performing the clustering on the respective generated signatures, thereby the created clusters include a collection of signatures respective of the multimedia data elements, wherein a signature is generated from multiple patches of a multimedia data element, wherein multiple patches are of random length and random position within the multimedia data element;
- and
- storing the first cluster in a storage unit.
An apparatus for unsupervised clustering of a large-scale collection of multimedia data elements, comprising:
- at least one processing unit;
- a storage unit for storing one or more clusters of multimedia data elements;
- a memory coupled to the at least one processing unit, of which at least a portion of the memory contains instructions that when executed by the at least one processing unit configure the apparatus to:
- generate a first cluster from the large-scale collection by:
- matching each of the multimedia data elements to all other multimedia data elements in the large-scale collection,
- determining a clustering score for each match being performed,
- clustering multimedia data elements having a clustering score above a threshold to create the first cluster;
- wherein the generation of the first cluster includes generating for each of the multimedia data elements at least one respective signature and performing the clustering on the respective generated signatures, thereby the created clusters include a collection of signatures respective of the multimedia data elements, wherein a signature is generated from multiple patches of a multimedia data element, wherein multiple patches are of random length and random position within the multimedia data element;
- and
- storing the first cluster in the storage unit.
Dependencies are inferred from the claim text.