An image semantic segmentation model and a segmentation method

By combining feature extraction, fusion, and deep supervised training modules, the applicability and efficiency of the image segmentation model are optimized, solving the problems of insufficient applicability and high computational cost of existing models in different fields, and achieving efficient image semantic segmentation.

CN116468740BActive Publication Date: 2026-05-05CHONGQING UNIV OF TECH
2 Cites 0 Cited by

Patent Information

Application Number
CN202310463484.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-26
Publication Date
2026-05-05
Estimated Expiration
2043-04-26

AI Technical Summary

Technical Problem

Existing image segmentation models are not applicable in different fields, especially when dealing with targets with large differences in shape distribution and size. In addition, the models are deep and computationally expensive, making it difficult to achieve a balance between time and space costs.

Method used

We employ a feature extraction module, a feature fusion module, and a deep supervised training module. Through a six-layer STDC backbone convolutional layer, a multi-scale spatial attention mechanism, deformable convolution, and deep supervised training methods, we optimize feature extraction and fusion to improve the model's generalization ability and segmentation performance.

Benefits of technology

It achieves robust performance across different segmentation tasks, improves the extraction capabilities of small targets and fuzzy boundary features, reduces computational cost and time complexity, and meets lightweight requirements.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application discloses an image semantic segmentation model which is composed of a designed mixed attention focusing method, an attention rectification residual module (ARRM) and a mixed feature integration module (MFFM), is trained in a deep supervision mode as a whole, is reasonably provided with deformable convolution, is combined with a constructed multi-scale spatial attention module (MSP) and a double-pooling attention module (DPA) to be jointly optimized, and the problem of small target feature segmentation difficulty is solved. In order to reflect the excellent performance of the model on a downstream task, the training between different data sets is completed in a transfer learning mode, and the application range of the model is expanded. Finally, grouping convolution is added in the backbone network, the calculation cost is greatly reduced, and the problem of high training cost of the segmentation model is solved under the premise of guaranteeing the segmentation effect.
Need to check novelty before this filing date? Find Prior Art