Neural Network Image Segmentation Independent Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network image segmentation methods require large datasets labeled consistently, which is resource-intensive and time-consuming, and often involve complex post-processing phases.
Innovation Solution
Training a neural network using datasets labeled independently from each other without a common taxonomy, where different classes are labeled differently across datasets, and using a loss function to evaluate and optimize neural network outputs, allowing pixels to be mapped to spaced-apart clusters in a multi-dimensional domain for segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current neural network image segmentation methods are used with consistently labeled datasets, then segmentation accuracy is improved, but training time and computing resources increase significantly
Solution Approach 1:
The patent changes the labeling parameter from consistent taxonomy to independent labeling across datasets. By allowing each dataset to be labeled independently without a common taxonomy, the system reduces the time and resources needed for data preparation while still achieving effective segmentation through the neural network's ability to learn from diverse labeling patterns.
Solution Approach 2:
The neural network is designed to handle multiple labeling schemes simultaneously. It processes pixels from datasets with different independent labelings and learns to map them to meaningful segments, making the system universal enough to work with various data sources without requiring uniform labeling conventions.
2Reliability
If large datasets with consistent labeling are used for training, then segmentation reliability is improved, but device complexity and resource requirements increase
Solution Approach 1:
The patent changes the data structure parameter from uniformly labeled datasets to independently labeled datasets. This allows the system to maintain reliability by processing diverse data sources while reducing the complexity of data preparation and management, as no complex consensus labeling process is needed.
Solution Approach 2:
The neural network performs self-service by automatically learning to reconcile different independent labelings during training. It develops its own understanding of segment boundaries without requiring external intervention to harmonize the labeling schemes, thereby reducing the need for complex preprocessing pipelines.
3Measurement precision
If traditional post-processing phases are applied to neural network results, then segmentation precision is improved, but processing time increases
Solution Approach 1:
The patent merges the segmentation decision-making process into the neural network training phase itself. By incorporating the segment identification logic directly into the network architecture and training objective, it eliminates the need for separate post-processing phases, thereby maintaining precision while significantly improving processing speed.
Solution Approach 2:
The neural network performs preliminary action by learning to directly output segment identifiers during training. Instead of generating raw predictions that require subsequent processing, the network is trained to produce final segmentation results directly, performing the segmentation decision in advance and eliminating later processing steps.
Data Source
AI summary
A method for image segmentation includes receiving, by a processing device, an image. The method further includes applying a machine learning model to the image, wherein the machine-learning model is trained by a training process comprising evaluating training outputs generated during the training process using a loss function. The method further includes obtaining, for each pixel of multiple pixels of the image, an output of the machine learning model within a multi-dimensional domain, wherein the output is obtained by providing the machine-learning model with pixels of different classes of segments of the image that are mapped to spaced apart clusters associated with different axes of the multi-dimensional domain. The method further includes determining, using the machine-learning model and for each pixel of multiple pixels of the image, a class of a segment that comprises the pixel by finding a closest axis to the output.