Metadata Computation for Multi-Type Media Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in efficiently analyzing unstructured media objects, such as videos, images, and textual data, due to the lack of effective metadata generation and management, particularly when dealing with multiple media types and large volumes of content.
Innovation Solution
A metadata computation apparatus and method that utilizes processors to compute and store metadata for multiple media types, including unstructured objects, by employing AI models and inference engines, and organizing this metadata in a unified database, facilitating efficient tagging and retrieval of target features across different media types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If metadata is computed for multiple media types using AI models, then the analysis capability of unstructured media objects is improved, but the device complexity and processing time increase
Solution Approach 1:
The storage controller is designed to perform multiple functions: it manages storage operations and simultaneously computes metadata for multiple media types (textual data, images, videos, sensor data) using AI inference engines. This multi-functionality allows the same hardware component to handle diverse media analysis tasks, improving analysis capability without proportionally increasing device complexity
Solution Approach 2:
The patent introduces metadata as an intermediary layer between raw unstructured media objects and analysis applications. The metadata computation apparatus processes media objects and generates structured metadata that simplifies subsequent analysis tasks, effectively mediating between complex media data and application requirements
2Loss of information
If metadata computation is performed for all media objects, then the information availability is improved, but the processing time and loss of time increase
Solution Approach 1:
The system computes metadata in advance during the storage process. The storage controller generates metadata for media objects as they are being stored or en-route to be stored, rather than processing them later when needed. This preliminary action ensures information is readily available when applications need it, eliminating subsequent processing delays
Solution Approach 2:
The metadata computation operates continuously in the background during storage operations. The storage controller maintains continuous processing of media objects, generating metadata streams that keep pace with data ingestion, ensuring information availability without interrupting the storage workflow
3Measurement precision
If AI models are used to generate metadata, then the measurement precision of target features is improved, but the use of energy and computational resources increase
Solution Approach 1:
The system extracts only the essential target features and relevant metadata from media objects using AI models, rather than processing and analyzing all aspects of the media data. This selective extraction approach maintains high detection precision for target features while reducing overall computational energy consumption
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Metadata computation apparatus (56) includes a host interface (68), a storage interface (60) and one or more processors (64). The host interface is configured to communicate over a computer network (28) with one or more remote hosts (24). The storage interface is configured to communicate with one or more non-volatile memories (52) of one or more storage devices (44). The processors are configured to manage local storage or retrieval of media objects in the non-volatile memories, to compute metadata for a plurality of media objects that are stored, or are en-route for storage, on the storage devices, wherein the media objects are of multiple media types, wherein the computed metadata tags a target feature in the media objects of at least two different media types among the multiple media types, and to store, in the non-volatile memories, the metadata tagging the target feature found in the at least two different media types, for use by the hosts.