相关实验视频

Updated: Jan 16, 2026

Robotized Testing of Camera Positions to Determine Ideal Configuration for Stereo 3D Visualization of Open-Heart Surgery

Robotized Testing of Camera Positions to Determine Ideal Configuration for Stereo 3D Visualization of Open-Heart Surgery

Published on: August 12, 2021

AGSK-Net:适应性几何感知立体KANformer网络,用于全球和本地无监督立体匹配

Qianglong Feng¹, Xiaofeng Wang¹, Zhenglin Lu¹

¹School of Mathematical and Physical Sciences, Chongqing University of Science and Technology, Chongqing 401331, China.

Sensors (Basel, Switzerland)

|September 27, 2025

概括

本研究介绍了自适应几何意识的立体KANformer网络 (AGSK-Net),用于无监督的立体匹配,通过结合几何先验和增强特征融合,提高在薄弱纹理和遮蔽等具有挑战性的领域的性能.

科学领域:

计算机视觉计算机视觉
深度学习 (Deep Learning) 是一种深度学习.
几何计算机视觉几何计算机视觉

背景情况:

无监督的立体声匹配对于3D重建至关重要,但与复杂的区域 (如薄弱的纹理和遮) 相斗争.
使用卷积神经网络 (CNN) 和视觉转换器 (ViT) 的现有方法由于局部受体场,缺乏几何先验和有限的表达性而存在局限性.

研究的目的:

开发一个先进的网络,即适应几何意识的立体KANformer网络 (AGSK-Net),以克服当前无监督立体相匹配技术的局限性.
加强对复杂图像区域的全球上下文和非线性表达力的建模.

主要方法:

拟议的自适应几何意识多头自我注意 (AG-MSA) 将极几何先验嵌入到ViT.
引入空间组-理性KAN (SGR-KAN) 取代MLP,整合理性函数以改善非线性表达.
开发了一个动态候选门融合 (DCGF) 模块,用于多个规模的全球和本地特征的自适应融合.

主要成果:

AGSK-Net在无监督立体声匹配中展示了最先进的准确性.
该网络在各种数据集中显示出卓越的概括性,包括Scene Flow,KITTI 2012/2015和Middlebury 2021.
在具有薄弱纹理和遮的具有挑战性的区域观察到显著的性能改善.

结论:

AGSK-Net有效地解决了以前无监督立体声匹配方法的局限性.
拟议的AG-MSA和SGR-KAN模块增强了几何意识和非线性建模能力.
该网络提供了一个强大的解决方案,用于准确和可泛化的无监督立体声匹配.

关键词:

这就是为什么KAN KAN KAN KAN.适应性融合适应性融合之前的极几何学之前的极几何学.斯温变压器斯温变压器无监督的立体声匹配.

更多相关视频

A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers

A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers

Published on: January 18, 2020

Medical-grade Sterilizable Target for Fluid-immersed Fetoscope Optical Distortion Calibration

Medical-grade Sterilizable Target for Fluid-immersed Fetoscope Optical Distortion Calibration

Published on: February 23, 2017

相关概念视频

Relative Motion Analysis using Rotating Axes-Problem Solving

Relative Motion Analysis using Rotating Axes-Problem Solving

Consider a crane whose telescopic boom rotates with an angular velocity of 0.04 rad/s and angular acceleration of 0.02 rad/s2. Along with the rotation, the boom also extends linearly with a uniform speed of 5 m/s. The extension of the boom is measured at point D, which is measured with respect to the fixed point C on the other end of the boom. For the given instant, the distance between points C and D is 60 meters.
Here, in order to determine the magnitude of velocity and acceleration for point...

Types of Global Positioning System Surveys

Types of Global Positioning System Surveys

GPS surveying methods vary in application, accuracy, and data collection techniques, catering to diverse surveying and mapping needs. Static GPS, kinematic GPS, and real-time kinematic (RTK) surveying are widely used. Each technique offers distinct advantages.Static GPS involves placing one receiver at a known reference point and another at the target point. It collects exact positional data by observing multiple satellite ranges over an extended period, achieving centimeter-level accuracy for...