一种基于改进HDNet的图像协调方法
By improving the HDNet model and combining the hollow spatial pooling pyramid structure and dual attention mechanism, the image coordination process is optimized, solving the problems of high computational cost and long training time, and improving the quality and harmony of image synthesis.
CN117496329BActive Publication Date: 2026-07-17CIVIL AVIATION UNIV OF CHINA
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CIVIL AVIATION UNIV OF CHINA
- Filing Date
- 2023-12-14
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing deep learning models are computationally intensive, time-consuming to train, and produce poor-quality synthesized images in image coordination tasks.
Method used
An improved HDNet model is adopted, which combines a feature fusion module with a dilated spatial pooling pyramid structure and a dual attention mechanism. Through multi-scale feature extraction and dynamic adjustment, combined with a mask-aware global dynamic module with deformable convolution and residual connections, the image coordination process is optimized.
Benefits of technology
It improves the quality and harmony of synthesized images, reduces computation and training time, and enhances the model's ability to capture image context information and global visual consistency.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN117496329B_ABST
Abstract
一种基于改进HDNet的图像协调方法。其包括选取公开数据集并进行预处理，分成训练集和测试集；构建基于改进HDNet的图像协调模型；使用训练集和测试集对模型进行训练及测试；将随机选取的真实合成图像输入模型进行协调处理获得协调后的合成图像等步骤。本发明效果：空洞空间池化金字塔结构通过不同采样率的空洞卷积从不同尺度提取输入特征，能有效地扩大感受野，捕捉图像中的上下文信息。双重注意力机制同时关注图像的空间和通道信息，动态地调整不同区域的重要性，可增强模型对图像中重要区域的关注度。将掩码感知全局动态模块中卷积核部分的传统卷积替换为可变形卷积，可以提高模型的性能。
Need to check novelty before this filing date? Find Prior Art