Hierarchical Neural Network Semantic Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Many semantic segmentation techniques exhibit inaccurate object classifications and object boundary definitions due to differences in knowledge extraction at various resolution scales within neural networks hierarchies.
Innovation Solution
Implementing cross-layer object class and object boundary processing by sharing and determining features between layers in a hierarchy of neural networks, allowing each layer to improve its object classifications and boundary definitions based on features from other layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a hierarchy of pre-trained CNNs is used to process image data at different resolution scales, then the system can extract features at multiple scales, but inaccurate object classifications and boundary definitions occur due to knowledge extraction differences across layers
Solution Approach 1:
The patent merges object classification features and object boundary features from multiple layers of the CNN hierarchy into a unified feature representation. By combining features from layers processed at different resolution scales, the system achieves both accurate object classifications and precise boundary definitions, resolving the contradiction between these two metrics.
Solution Approach 2:
The patent introduces an intermediary processing mechanism that facilitates knowledge transfer between layers of the hierarchy. This intermediary process allows features to be shared and refined across different resolution scales, enabling accurate extraction of both object class information and boundary information that would otherwise be conflicting.
2Adaptability or versatility
If layers in the hierarchy process image data at different resolution scales, then multi-scale feature extraction is achieved, but knowledge extraction differences cause segmentation inaccuracies
Solution Approach 1:
The patent segments the feature extraction process into distinct object classification feature extraction and object boundary feature extraction pathways. By separating these functions and then integrating their results, the system maintains the benefits of multi-scale processing while achieving accurate segmentation through coordinated feature combination.
Solution Approach 2:
The patent creates a universal feature representation that serves multiple purposes: it captures both object class information and object boundary information from the same multi-scale hierarchical processing. This universal feature set enables the system to maintain adaptability across scales while achieving precision in segmentation tasks.
Data Source
AI summary
Aspects described herein relate to various methods, systems and apparatuses that may improve the accuracy of object classifications and object boundary definitions for a semantic segmentation technique. For example, the semantic segmentation technique may be based on a hierarchy of two or more layers. The two or more layers may include neural networks that analyze image data at different resolution scales. Each layer of the hierarchy may determine object boundary features and object class features. Each layer of the hierarchy may share its object boundary features and/or its object class features with one or more other layers in the hierarchy. In turn, each of other layers of the hierarchy may determine its object boundary features and/or its object class features based on the shared features.