3D Model Authentication via Visual Descriptor Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online virtual experience platforms face challenges in accurately classifying and authenticating users and 3D models due to subjective descriptions and potential mislabeling, which affects machine learning and content filtering.
Innovation Solution
A method that generates 2D images from 3D models, creates descriptors based on image recognition, and allows user selection to authenticate and modify weights for AI training data, enhancing classification and user verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If subjective text descriptions are used for 3D models, then ease of operation is improved, but measurement precision and classification accuracy deteriorate
Solution Approach 1:
The system generates multiple 2D images from the 3D model and creates corresponding visual descriptors that objectively represent the model's appearance. These visual descriptors serve as accurate copies of the model's visual properties, replacing subjective text descriptions while maintaining ease of use through automated processing.
Solution Approach 2:
The system replaces manual text labeling with automated image recognition algorithms that generate descriptors based on visual analysis of 2D images. This substitution of mechanical manual labeling with automated computational analysis improves classification accuracy while maintaining operational simplicity.
2Measurement precision
If automated image recognition is used to generate descriptors, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system segments the complex task of 3D model classification into multiple manageable steps: generating 2D images from the 3D model, extracting visual descriptors from these images, and using these descriptors for classification. This segmentation reduces the complexity of any single component while improving overall precision.
Solution Approach 2:
The system introduces 2D images and visual descriptors as intermediary representations between the 3D model and the classification system. These intermediaries simplify the relationship between complex 3D data and classification algorithms, reducing overall system complexity while maintaining high measurement precision.
3Productivity
If user authentication is integrated with training data capture, then productivity is improved, but device complexity increases
Solution Approach 1:
The system performs multiple functions through a single integrated process: generating 2D images from 3D models, creating visual descriptors, authenticating users based on their selections, and capturing training data. This multi-functionality improves productivity by eliminating separate processes while the modular design keeps complexity manageable.
Solution Approach 2:
The system automatically captures training data as part of the normal user authentication process, allowing the system to serve itself by using user interactions for both authentication and data collection purposes. This self-service approach improves productivity without requiring separate complex data collection mechanisms.
4Reliability
If multiple descriptors are presented for user selection, then reliability of authentication is improved, but ease of operation deteriorates
Solution Approach 1:
The system presents multiple descriptors (potentially more than needed) for user selection, allowing users to confirm their understanding of the 3D model by selecting appropriate descriptors. This partial action approach improves authentication reliability while the automated processing of selections maintains ease of operation.
Data Source
AI summary
Some implementations relate to methods, systems, and computer-readable media for automatically capturing training data. In some implementations, a computer-implemented method includes generating a respective set of descriptors based on content identified in a 3D model, presenting, in a user interface, the 3D model and two or more descriptors, receiving, from the user interface, at least one selection from the two or more descriptors, authenticating a user based on the at least one selection, and modifying a weight of at least one descriptor associated with the at least one selection for an associated 2D image in a labeled training set for an artificial intelligence (AI) model.