Search research articles

关于 JoVE

概览领导团队博客 JoVE 帮助中心

作者

出版流程编辑委员会范围与政策同行评审常见问题投稿

图书馆员

用户评价订阅访问资源图书馆顾问委员会常见问题

研究

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments 存档

教育

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual 教师资源中心教师网站

使用条款与条件

Search research articles

相关实验视频

Updated: Sep 9, 2025

Deep Neural Networks for Image-Based Dietary Assessment

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

GIMS:基于自适应图形构造和图形神经网络的图像匹配系统

Xianfeng Song¹, Yi Zou¹, Zheng Shi¹

¹School of Microelectronics, South China University of Technology, Guangdong, China.

Neural networks : the official journal of the International Neural Network Society

|September 4, 2025

概括

本研究介绍了基于特征的图像匹配的自适应图形构造方法,使用图形神经网络 (GNN) 和变压器. 这种新方法显著提高了匹配性能和效率.

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序

Same author

Leveraging direct learning and reinforcement learning architectures for pre-equalization in high-speed VLC systems.

Optics express·2026

Same author

Development and application of an improved constitutive model for acidizing coal around fractured boreholes.

Scientific reports·2025

Same author

Centrifugal microfluidic chip with an air gap for oil-free production of enhanced adipogenic multicellular microspheres.

Lab on a chip·2025

Same author

Research on the Bearing Remaining Useful Life Prediction Method Based on Optimized BiLSTM.

Sensors (Basel, Switzerland)·2025

Same author

Targeted Isolation of ω-3 Polyunsaturated Fatty Acids from the Marine Dinoflagellate <i>Prorocentrum lima</i> Using DeepSAT and LC-MS/MS and Their High Activity in Promoting Microglial Functions.

Marine drugs·2025

Same author

Association of grey matter cerebral blood flow with white matter integrity in relation to youth bipolar disorder.

Journal of psychiatry & neuroscience : JPN·2025

科学领域:

计算机视觉
机器学习
图形神经网络

背景情况:

基于特征的图像匹配对于计算机视觉任务至关重要.
图形神经网络 (GNN) 与传统的关键点表示方法相比,具有更高的性能.
现有的图像匹配GN方法在图形构造和特征表示方面面临挑战.

研究的目的:

开发一种创新的自适应图形构造方法,以增强图像匹配.
将GNN与变压器集成,以改善空间和特征信息的表示.
在基于特征的图像匹配中实现最先进的性能.

主要方法:

引入了使用距离和动态值相似性过的自适应图构造.
结合图形神经网络 (GNN) 与变压器用于混合特征表示.
使用Sinkhorn算法进行最佳匹配和多GPU加速训练.

主要成果:

在整体匹配性能方面平均提高了3.8×-40.3×.
证明了自适应图形结构和混合模型的有效性.
通过比较实验验证了系统在广泛的图像数据集上的性能.

结论:

关键词:

图表匹配图表神经网络图像匹配机器学习

更多相关视频

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

相关实验视频

Last Updated: Sep 9, 2025

Deep Neural Networks for Image-Based Dietary Assessment

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

拟议的GNN-变压器混合模型具有自适应图形结构,可显著提升基于特征的图像匹配.
这种方法提供了强大而精确的图形结构,提高了匹配精度.
通过多GPU技术实现的高效培训,解决计算需求.