Voxel-Limited 3D Shape Completion With Intermediate Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing voxel-based three-dimensional shape completing methods, such as those described in Non-Patent Literature 1, have limitations in expressing spatial information effectively.
Innovation Solution
An information processing apparatus and method that includes processes for generating three-dimensional structure data, sampling, completing incomplete regions using a completion model, and training the model with a loss value to improve spatial expression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voxel-based three-dimensional shape completing method is used, then the completing process can be performed, but the ability of expressing space is limitative
Solution Approach 1:
The patent transforms the completion model from a voxel-based discrete space representation to a continuous space representation using coordinate transformation. This parameter change allows the model to express spatial information more flexibly and accurately, resolving the limitation of voxel-based methods while maintaining the completing process functionality.
Solution Approach 2:
The patent replaces the traditional voxel-based mechanical grid system with a neural network-based continuous coordinate system. This substitution enables more accurate and flexible spatial expression by eliminating the discrete voxel constraints while preserving the core completing function through learned transformations.
2Adaptability or versatility
If intermediate feature value is used in shape estimating process, then the spatial expression ability is improved, but the training complexity increases
Solution Approach 1:
The patent introduces intermediate feature values as mediators between the input three-dimensional data and the final completion output. These intermediate features capture spatial relationships and are used to guide the completion process, improving spatial expression while the modular training approach manages the associated complexity through staged optimization.
Solution Approach 2:
The patent segments the completion process into distinct stages: feature extraction, intermediate feature computation, and final completion. This segmentation allows the model to process spatial information in manageable steps, improving expression ability while making the training complexity tractable through modular optimization.
Data Source
AI summary
An information processing apparatus includes at least one processor, the at least one processor executing: an obtaining process of obtaining input data; a three-dimensional structure data generating process of generating three-dimensional structure data from the input data; a sampling process of generating sampled three-dimensional structure data by sampling the three-dimensional structure data; a completing process with respect to the sampled three-dimensional structure data with use of a completion model; an estimating process of executing a shape estimating process in which an intermediate feature value in the completing process is referred to; and a first training process of training the completion model with reference to a first loss value which is a loss value pertaining to a shape that has been obtained by the shape estimating process.