Image Segmentation Model Category Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Category unbalance in image processing tasks, where a large difference in ratios of samples of different categories, makes it challenging to accurately identify and segment unbalanced categories, such as small objects in images.
Innovation Solution
A method for image segmentation that involves selecting a to-be-discarded category based on its recall rate and processing the image based on the remaining categories, thereby reducing the amount of data to be processed and improving segmentation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all categories are processed in image segmentation, then segmentation completeness is improved, but computing requirements and processing time increase
Solution Approach 1:
The patent extracts and identifies the minority category (to-be-discarded category) from the image based on recall rate analysis, then processes only the remaining majority categories. This extraction approach allows the system to handle computationally intensive minority categories selectively while maintaining overall segmentation completeness, thereby resolving the contradiction between processing all categories and processing efficiency.
Solution Approach 2:
The patent changes the processing parameter by using recall rate as a threshold to determine which categories to process. By adjusting the recall rate parameter, the system can dynamically control the balance between segmentation completeness and processing efficiency, allowing flexible adaptation to different computing resource constraints while maintaining reliable segmentation results.
2Measurement precision
If recall rate is increased for unbalanced categories, then segmentation accuracy is improved, but the amount of data to be processed increases
Solution Approach 1:
The patent performs preliminary analysis of recall rates for different categories before actual segmentation processing. By pre-calculating which categories have low recall rates and identifying them as to-be-discarded categories, the system can focus processing resources on categories that need improvement, thereby improving segmentation accuracy for unbalanced categories without unnecessarily increasing the overall data processing volume.
Solution Approach 2:
The patent applies different processing strategies to different categories based on their local characteristics (recall rates). Instead of uniformly processing all categories with the same computational resources, the system selectively enhances processing for specific unbalanced categories identified by low recall rates, thereby improving segmentation accuracy where needed while avoiding unnecessary processing of already well-segmented categories.
3Measurement precision
If category weights are adjusted during training, then segmentation accuracy for unbalanced categories is improved, but training complexity increases
Solution Approach 1:
The patent changes training parameters by dynamically adjusting category weights based on recall rate analysis. The system identifies unbalanced categories and assigns them different weight values during training, allowing the model to focus more attention on difficult-to-segment categories. This parameter adjustment approach improves segmentation accuracy for unbalanced categories while maintaining relatively simple training procedures through automated weight calculation.
Data Source
AI summary
Embodiments of the present disclosure relate to a method, an electronic device, and a computer program product for image segmentation. The method may be executed by a trained image segmentation model. The method includes: obtaining a to-be-processed image, wherein the to-be-processed image includes objects of a plurality of categories. The method further includes: selecting a to-be-discarded category in the to-be-processed image according to a recall rate of each of the plurality of categories obtained in advance. The method further includes: processing the to-be-processed image based on a plurality of remaining categories in the plurality of categories other than the to-be-discarded category to obtain a segmented image. Through the method, computing resources required for a segmentation processing task can be greatly reduced, a processing amount of image data can be reduced, and an image processing speed can be increased.