3D Shape Model Fusion Using Likelihood-Based Subregion Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D shape reconstruction techniques using deep learning often result in a loss of detail when creating a single 3D shape model from averaged observational information, leading to inferior estimation of unobserved regions.
Innovation Solution
Obtain multiple 3D shape models for the same object from different viewpoints, calculate likelihoods for subregions within these models, and combine subregions with the highest likelihoods to create a detailed 3D shape model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single 3D shape model is estimated from averaged observational information of the entire sequence, then the processing efficiency is improved, but the level of detail in the reconstructed model deteriorates
Solution Approach 1:
The patent divides the 3D space into multiple subregions (e.g., front, back, left, right sides) and processes each subregion separately using the most appropriate observational information. This segmentation allows the system to maintain high processing efficiency while preserving detailed information in each specific region, resolving the contradiction between overall efficiency and local detail quality.
2Manufacturing precision
If multiple different 3D shape models are created and combined, then the level of detail is improved, but the device complexity increases
Solution Approach 1:
The patent applies local quality by evaluating and selecting the most appropriate 3D shape model for each specific subregion based on likelihood calculations. Instead of uniformly processing all regions with the same complexity, the system adapts the processing approach to each region's characteristics, improving detail where needed while managing overall system complexity.
Solution Approach 2:
The patent replaces complex manual or mechanical model combination processes with automated likelihood-based selection and neural network-based rendering. This substitution reduces the practical complexity of handling multiple models while maintaining the ability to produce high-detail reconstructions.
3Device complexity
If a single 3D shape model is created from averaged information, then the device complexity is reduced, but the measurement precision of unobserved regions deteriorates
Solution Approach 1:
The patent performs preliminary actions by generating multiple candidate 3D shape models from different observational perspectives before final reconstruction. These pre-generated models serve as candidates that are later evaluated and combined, allowing the system to maintain low complexity while improving measurement precision through informed selection rather than simple averaging.
Data Source
AI summary
A technique disclosed herein creates a 3D shape model with a higher level of detail based on multiple different 3D shape models. Multiple 3D shape models for the same object are obtained and the likelihood (first likelihood) of each of polygons constituting each of the obtained 3D shape models is calculated. Using estimated skeleton data, the multiple 3D shape models with different poses are transformed into the same pose. Then, scale conversion is performed on the pose-transformed 3D shape models containing the pose-transformed skeleton data. The likelihood (second likelihood) of polygons in each certain subregion is calculated from the calculated likelihoods (first likelihoods) of the polygons. Subregions with the highest likelihoods (second likelihoods) of the polygons are extracted from the multiple 3D shape models, and combined to create a single 3D shape model.