Metadata generation for multiple object types

1. Metadata computation apparatus, comprising:

a host interface, configured to communicate over a computer network with one or more remote hosts;

a storage interface, configured to communicate with one or more non-volatile memories of one or more storage devices; and

one or more processors, configured to:

manage local storage or retrieval of media objects in the one or more non-volatile memories;

compute metadata for a plurality of the media objects that are stored, or that are en-route to be stored, on the one or more storage devices, wherein the media objects are of multiple media types, at least some of the media objects being unstructured media objects, and wherein the computed metadata comprises a common identifier assigned to locations at which a same common target feature appears in both (i) first unstructured media objects of a first media type and (ii) second unstructured media objects of a second media type, different from the first media type, the locations computed in accordance with different location metrics that are respectively defined for the respective media types; and

store, in the one or more non-volatile memories, the metadata comprising the common identifier assigned to the common target feature found in the at least two different media types, for use by the one or more hosts.

2. The metadata computation apparatus according to claim 1 , wherein, for a media item that comprises a sequence of frames, the one or more processors are configured to identify and assign the common identifier to one or more of the frames in which the common target feature appears.

3. The metadata computation apparatus according to claim 1 , wherein, for a media item that comprises at least a frame, the one or more processors are configured to identify and assign the common identifier to one or more coordinates in the frame in which the common target feature appears.

4. The metadata computation apparatus according to claim 1 , wherein the one or more processors are configured to receive from the one or more hosts, over the computer network, one or more models that specify extraction of the metadata from the media objects, and to generate the metadata based on the received models.

5. The metadata computation apparatus according to claim 4 , wherein the one or more processors are configured to receive from the one or more hosts a respective model for each of the multiple media types.

6. The metadata computation apparatus according to claim 4 , wherein the one or more processors are configured to receive, as the one or more models, one or more pre-trained Artificial Intelligence (AI) models.

7. The metadata computation apparatus according to claim 6 , wherein the one or more processors are configured to generate the metadata by applying a same AI inference engine to the AI models.

8. The metadata computation apparatus according to claim 1 , wherein the one or more processors are configured to organize the media objects in multiple batches corresponding to the media types, and to compute the metadata over each of the batches.

9. The metadata computation apparatus according to claim 1 , wherein the one or more processors are configured to generate the metadata during idle periods during which at least some resources of the one or more processors are free from managing storage of the media objects.

10. The metadata computation apparatus according to claim 1 , wherein the one or more processors are configured to combine the metadata, which assigns the common identifier to the common target feature, in a unified metadata database that identifies at least one attribute selected from a group of attributes consisting of a media type, a file identifier of a file containing the media object, and a location of the media object within the file.

11. Metadata computation apparatus, comprising:

a host interface, configured to communicate over a computer network with one or more remote hosts;

a storage interface, configured to communicate with one or more non-volatile memories of one or more storage devices; and

one or more processors, configured to:

manage local storage or retrieval of media objects in the one or more non-volatile memories;

compute metadata for a plurality of the media objects that are stored, or that are en-route to be stored, on the one or more storage devices, wherein the media objects are of multiple media types, and wherein the computed metadata comprises a common identifier assigned to occurrences of a same common target feature in both (i) first media objects of a first media type and (ii) second media objects of a second media type, different from the first media type; and

store, in the one or more non-volatile memories, the metadata comprising the common identifier assigned to the common target feature found in the at least two different media types, for use by the one or more hosts, including combining, in a unified metadata database, metadata that comprises the common identifier assigned to the common target feature and that was extracted from different media sources or extracted by different processors.

12. A method for metadata computation, the method comprising:

communicating by a storage controller of one or more storage devices over a computer network with one or more remote hosts, and communicating with one or more non-volatile memories of the one or more storage devices;

using the storage controller, managing local storage or retrieval of media objects in the one or more non-volatile memories, computing metadata for a plurality of the media objects that are stored, or that are en-route to be stored, on the one or more storage devices, wherein the media objects are of multiple media types, at least some of the media objects being unstructured media objects, and wherein the computed metadata comprises a common identifier assigned to locations at which a same common target feature appears in both (i) first unstructured media objects of a first media type and (ii) second unstructured media objects of a second media type, different from the first media type, the locations computed in accordance with different location metrics that are respectively defined for the respective media types; and

storing the metadata, which comprises the common identifier assigned to the common target feature found in the at least two different media types, in the one or more non-volatile memories for use by the one or more hosts.

13. The method for metadata computation according to claim 12 , wherein, for a media item that comprises a sequence of frames, assigning the common identifier to the locations comprises identifying and assigning the common identifier to one or more of the frames in which the common target feature appears.

14. The method for metadata computation according to claim 12 , wherein, for a media item that comprises at least a frame, assigning the common identifier to the locations comprises identifying and assigning the common identifier to one or more coordinates in the frame in which the common target feature appears.

15. The method for metadata computation according to claim 12 , comprising receiving from the one or more hosts, over the computer network, one or more models that specify extraction of the metadata from the media objects, and generating the metadata based on the received models.

16. The method for metadata computation according to claim 15 , wherein receiving the models comprises receiving from the one or more hosts a respective model for each of the media types.

17. The method for metadata computation according to claim 15 , wherein receiving the models comprises receiving one or more pre-trained Artificial Intelligence (AI) models.

18. The method for metadata computation according to claim 17 , wherein computing the metadata comprises applying a same AI inference engine to the AI models.

19. The method for metadata computation according to claim 12 , wherein computing the metadata comprises organizing the media objects in multiple batches corresponding to the media types, and computing the metadata over each of the batches.

20. The method for metadata computation according to claim 12 , wherein computing the metadata comprises generating the metadata during idle periods during which at least some resources of the storage controller are free from managing storage of the media objects.

21. The method for metadata computation according to claim 12 , wherein storing the metadata comprises combining the metadata, which comprises the common identifier assigned to the common target feature, in a unified metadata database that identifies at least one attribute selected from a group of attributes consisting of a media type, a file identifier of a file containing the media object, and a location of the media object within the file.

22. A method for metadata computation, the method comprising:

communicating by a storage controller of one or more storage devices over a computer network with one or more remote hosts, and communicating with one or more non-volatile memories of the one or more storage devices;

using the storage controller, managing local storage or retrieval of media objects in the one or more non-volatile memories, computing metadata for a plurality of the media objects that are stored, or that are en-route to be stored, on the one or more storage devices, wherein the media objects are of multiple media types, and wherein the computed metadata comprises a common identifier assigned to occurrences of a same common target feature in both (i) first media objects of a first media type and (ii) second media objects of a second media type, different from the first media type; and

storing the metadata, which comprises the common identifier assigned to the common target feature found in the at least two different media types, in the one or more non-volatile memories for use by the one or more hosts, including combining, in a unified metadata database, metadata that comprises the common identifier assigned to the common target feature and that was extracted from different media sources or extracted by different processors.