3D Pose Estimation Using Dynamic Graph Matrix Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D pose estimation methods using graph convolution networks face challenges in processing distorted inputs due to occlusion or noise, leading to degraded performance, especially in maintaining memory usage and inference time.
Innovation Solution
A 3D pose estimation apparatus that generates a dynamic graph matrix by fusing an affinity matrix with a predefined static graph matrix, using methods such as multiplication or summation modulation, and applying regular symmetric constraints to improve pose estimation performance while maintaining memory and inference time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a static graph matrix is used in the GCN layer, then the model structure is simple and inference time is fast, but the processing capability for distorted input data (occlusion or noise) is degraded
Solution Approach 1:
The patent applies dynamics by transforming the static graph matrix into a dynamic graph matrix that can adapt to different input conditions. The graph matrix is no longer fixed but changes based on the input feature vector, allowing the model to handle distorted inputs effectively while maintaining a relatively simple overall structure.
Solution Approach 2:
The patent changes the parameters of the graph matrix dynamically based on the input data. By using the input feature vector to modulate the graph matrix parameters, the model can adapt to various data cases including occlusion and noise, improving reliability without requiring a completely complex new architecture.
2Reliability
If a dynamic graph matrix is generated by fusing affinity matrix with static graph matrix, then the processing capability for various data cases is improved, but the memory usage and computational complexity increase
Solution Approach 1:
The patent merges the affinity matrix (which captures input-specific relationships) with the static graph matrix (which provides structural priors) to create a dynamic graph matrix. This combination allows the model to leverage both data-driven and structure-driven information, improving processing capability while keeping the computational complexity manageable through efficient fusion operations.
Solution Approach 2:
The dynamic graph matrix acts as an intermediary that bridges the static graph structure and the input-specific affinity relationships. This intermediary enables flexible processing of various data cases by translating between the fixed structural constraints and the variable input characteristics without requiring direct complex interactions between all components.
3Adaptability or versatility
If the graph matrix is replaced dynamically based on input features, then the adaptability to different data cases is improved, but the inference time increases
Solution Approach 1:
The patent implements parameter changes by dynamically adjusting the graph matrix parameters based on input features. This allows the model to adapt to different data cases (occlusion, noise, various poses) by changing the graph connectivity and weights on-the-fly, achieving high adaptability while maintaining reasonable inference time through efficient parameter modulation techniques.
Data Source
AI summary
Provided is a pose estimation method in a 3D pose estimation apparatus for estimating a 3D pose of an object based on a graph convolution network (GCN). The pose estimation method comprises inputting a feature vector for a joint of the object; generating an affinity matrix according to the feature vector; generating a dynamic graph matrix by fusing the affinity matrix with a predefined static graph matrix of the graph convolution network; and estimating a 3D pose of the object for the feature vector by replacing the static graph matrix of the graph convolution network with the dynamic graph matrix. As a result, the performance of estimating a 3D pose can be greatly improved while almost maintaining the memory usage and inference time of the existing graph convolution network (GCN).