3D Model Reconstruction from 2D Images via Neural Network Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital literature representations using multiple 2D views of a 3D object are disjointed, making it difficult for readers to accurately reconstruct the original 3D object in their minds.
Innovation Solution
A method and apparatus that utilize a neural network for information fusion and 3D model reconstruction by inputting 2D images from multiple viewing angles, allowing for the generation of a 3D model from these images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If multiple 2D views are used to describe a 3D object, then the completeness of object information is improved, but the ease of mental reconstruction deteriorates
Solution Approach 1:
The patent merges multiple separate 2D views into a unified 3D model that integrates all viewing angle information. The 3D model consolidation module combines the plurality of 2D images from different angles into a single coherent 3D representation, allowing readers to perceive the complete object structure without mentally assembling separate views.
Solution Approach 2:
The patent transitions from 2D representations to a 3D representation by adding the depth dimension. The neural network processes 2D images from multiple angles and reconstructs them into a 3D model, enabling readers to view the object from any angle interactively, thus solving the limitation of fixed-angle 2D views.
2Measurement precision
If multiple 2D views are provided, then the accuracy of object representation is improved, but the device complexity increases
Solution Approach 1:
The patent replaces complex manual processes of creating and managing multiple 2D views with an automated neural network system. The 3D reconstruction neural network automatically processes 2D images from various angles and generates accurate 3D models, eliminating the need for manual 3D modeling while maintaining high representation accuracy.
Solution Approach 2:
The patent creates a multi-functional system where a single 3D model serves multiple purposes: it can be viewed from any angle, rotated interactively, and provides complete object information. This unified 3D representation replaces the need for multiple separate 2D views, simplifying the system while improving accuracy.
3Loss of information
If 3D model reconstruction is implemented, then the user understanding of content is improved, but the processing time increases
Solution Approach 1:
The patent performs 3D model reconstruction in advance during the document processing stage. The 3D model is generated beforehand and stored, so when readers access the document, they can immediately interact with the pre-processed 3D model without experiencing processing delays. This preliminary action separates the time-consuming reconstruction process from the user viewing experience.
Data Source
AI summary
Disclosed are a method, an apparatus, a device and a storage medium for reconstructing a 3D model from 2D images, comprising: obtaining two-dimensional images respectively corresponding to at least two viewing angles of a three-dimensional object; and inputting the two-dimensional images respectively corresponding to the at least two viewing angles into a set neural network for information fusion and 3D model reconstruction so as to obtain a 3D model of the three-dimensional object.