3D Model Search via Image Vector Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search methods for three-dimensional models are inadequate when users or algorithms use different terms or languages to describe models, leading to insufficient or irrelevant search results.
Innovation Solution
A hybrid machine-learning model that generates a representation vector from a two-dimensional image, using supervised and unsupervised learning layers to identify and search for similar three-dimensional models, allowing for effective querying and comparison across different descriptive frameworks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If text-based search terms are used to query three-dimensional models, then the search can be performed using simple keywords, but the search results are inadequate when users use different terms or languages to describe the same model
Solution Approach 1:
The patent introduces an intermediary representation system (embeddings, vector spaces, and machine learning models) that mediates between user queries and model databases. This intermediary layer translates different descriptive frameworks into a common semantic space, enabling accurate matching regardless of the specific terms or languages used by users.
Solution Approach 2:
The system transforms search queries from text-based parameters to numerical vector representations through machine learning models. By changing the parameter space from discrete text tokens to continuous vector embeddings, the system enables semantic similarity search that is insensitive to specific word choices or languages.
2Measurement precision
If two-dimensional images are converted into high-dimensional representation vectors, then accurate model identification is achieved, but computational resources increase
Solution Approach 1:
The patent pre-computes and stores embedding representations for all three-dimensional models in the database before query time. This preliminary action allows the system to avoid performing complex high-dimensional transformations during actual search operations, reducing real-time computational resource consumption while maintaining high matching accuracy.
3Measurement precision
If a hybrid machine-learning model with multiple layers is used to generate representation vectors, then accurate feature extraction is achieved, but the system complexity increases
Solution Approach 1:
The patent divides the machine-learning model into distinct functional layers (supervised learning layers for feature extraction and unsupervised learning layers for representation learning). This segmentation allows each layer to specialize in specific tasks, improving overall accuracy while enabling modular training and deployment that manages system complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
One or more three-dimensional models that corresponds to at least one two-dimensional image are determined by receiving image data that corresponds to the at least one two-dimensional image, generating features based on the image data corresponding to the at least one two-dimensional image, generating a representation vector for the at least one two-dimensional image by transforming the features into a predetermined amount of numerical representations corresponding to the features, and outputting the representation vector for the at least one two-dimensional image to facilitate a search query for the one or more three-dimensional models associated with the at least one two-dimensional image.