Object-Level Relational Index for Image Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image-based search technologies are limited in their ability to identify and relate individual objects within images, failing to provide efficient retrieval of relevant information at a per-object level of granularity.
Innovation Solution
A computer-implemented technique using a machine-trained model to create a relational index of objects within images, where each object is associated with attributes such as latent semantic vectors, allowing for the identification of related objects through their co-occurrence in images, thereby enabling efficient search and retrieval of complementary or similar items.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single composite image is used for search queries, then the search engine can identify candidate images, but it cannot retrieve information at a per-object level of granularity
Solution Approach 1:
The patent segments a composite image into multiple individual objects, creating separate index entries for each object rather than treating the image as a single unit. This segmentation enables per-object level search and retrieval, allowing users to query for specific objects within images and receive results at the object granularity level, directly resolving the contradiction between measurement precision and device complexity.
2Productivity
If traditional image-based search is used, then candidate images can be identified, but relevant information cannot be quickly extracted from large image corpora
Solution Approach 1:
The patent introduces an object relationship graph as an intermediary data structure that captures relationships between objects across multiple images. This graph serves as a mediator between the raw image corpus and the search query, enabling efficient extraction of relevant information by leveraging pre-computed object relationships rather than analyzing entire image corpora during query processing, thus improving productivity while preserving information.
3Adaptability or versatility
If objects are identified within images, then per-object search becomes possible, but the complexity of creating and managing relational indices increases
Solution Approach 1:
The patent transitions from traditional two-dimensional image space to a multi-dimensional object relationship space by creating relational indices that capture objects, their attributes, and relationships across multiple images. This dimensional transformation enables versatile per-object search capabilities while managing complexity through structured representation of object relationships in a separate semantic space rather than within the image data itself.
Data Source
AI summary
A computer-implemented technique is described herein for using a machine-trained model to identify individual objects within images. The technique then creates a relational index for the identified objects. That is, each index entry in the relational index is associated with a given object, and includes a set of attributes pertaining to the given object. One such attribute identifies at least one latent semantic vector associated with the given object. Each attribute provides a way of linking the given object to one or more other objects in the relational index. In one application of this technique, a user may submit a query that specifies a query object. The technique consults the relational index to find one or more objects that are related to the query object. In some cases, the query object and each of the other objects have a complementary relationship.