非线性多头交叉注意网络和可编程梯度信息,用于视线估计
Yujie Li1,2, Yuhang Hong1, Ziwen Wang1
1School of Artificial Intelligence, Guilin University of Eletronic Technology, Guilin, 541004, Guangxi, China.
Scientific reports
|July 27, 2025
概括
这项研究引入了一种新的目光估计方法,MCA-PGI,将CNN和变压器结合起来,以获得更好的准确性. 这种新的方法显著提高了基准数据集的性能.
科学领域:
- 计算机视觉 计算机视觉
- 机器学习 机器学习
- 人与计算机的交互
背景情况:
- 凝视估计对于人类行为分析和协助至关重要.
- 目前使用卷积神经网络 (CNN) 和注意力转换器的方法在特征提取和利用方面存在局限性.
- CNN捕捉有限的局部环境,而注意力机制低利用多尺度特征.
研究的目的:
- 提出一种具有可编程梯度信息 (MCA-PGI) 的新型非线性多头交叉注意网络,以提高视线估计.
- 为了克服现有的CNN和基于变压器的目光估计方法的局限性.
- 为了提高视线估计系统的准确性和稳定性.
主要方法:
- 开发了一种具有可编程梯度信息 (MCA-PGI) 的新型非线性多头交叉注意网络.
- 集成的CNN和变压器优势,利用辅助分支进行梯度信息.
- 采用非线性多头交叉注意力来融合全球视觉和多尺度混合特征.
主要成果:
- 与最先进的方法相比,MCA-PGI表现出强大的竞争力.
- 实现了显著的性能改进:MPIIFaceGaze上的2.5%和Eyediap数据集上的10.2%.
- 拟议的方法有效地保留了原始信息,并融合了多层次特征.
结论:
- 通过有效地结合本地和全球特征提取,MCA-PGI网络为目光估计提供了卓越的方法.
- 该方法显示了需要精确的人类行为分析的应用程序的巨大潜力.
- 开源实现可用于进一步的研究和开发.
相关概念视频
01:24One-Degree-of-Freedom System
In mechanical engineering, one-degree-of-freedom systems form the basis of a wide range of electrical and mechanical components. Using these models, engineers can predict the behavior of various parts in a larger system, which gives them insight into how different forces interact with each other.
A one-degree-of-freedom system is defined by an independent variable that determines its state and behavior. One example of a one-degree-of-freedom system is a simple harmonic oscillator, such as a...
A one-degree-of-freedom system is defined by an independent variable that determines its state and behavior. One example of a one-degree-of-freedom system is a simple harmonic oscillator, such as a...
01:27Curvilinear Motion: Normal and Tangential Components
When a car traverses a curved road, its motion can be elucidated by breaking it down into tangential and normal components. The car-centric coordinates attached to the vehicle move with it.
The positive direction of the t-axis aligns with the increasing position of the car along the curved path, denoted by the unit vector ut. Simultaneously, the n-axis, perpendicular to the t-axis, dissects the curved path into differential arc segments, each forming the arc of a circle with a radius of...
The positive direction of the t-axis aligns with the increasing position of the car along the curved path, denoted by the unit vector ut. Simultaneously, the n-axis, perpendicular to the t-axis, dissects the curved path into differential arc segments, each forming the arc of a circle with a radius of...
01:12Differential Leveling
Differential leveling is a precise method in surveying used to determine the elevation difference between two points. Its primary goal is to establish accurate vertical measurements to create level surfaces or grade lines critical for designing and constructing infrastructures such as roads, bridges, and buildings.The procedure for differential leveling begins with setting up and leveling the instrument at a point where the benchmark can be seen. The level rod is held on the benchmark (BM), and...
01:25Design Example: Traverse Angle Computations
Traverse angle computations are a critical component of surveying, used to compute the internal angles within a closed traverse. A traverse consists of a series of connected lines forming a closed loop, often used for land boundary delineation or mapping. Calculating the internal angles ensures accuracy in the traverse geometry and is essential for checking survey data integrity.The process begins with known azimuths and bearings of the traverse sides. Internal angles at each vertex are...
01:26Linearization and Approximation
Linearization is a mathematical technique used to approximate complex, nonlinear functions with simpler linear models in the vicinity of a chosen reference point. The method is based on the idea that, although a function may be difficult to evaluate exactly, its behavior near a specific input value can often be closely approximated by the tangent line at that point. This approach is particularly useful when small deviations from a known value are involved.Consider the square root function, for...
01:29Application of Linearization and Approximation
A drone flying through complex terrain often relies on more than one sensing method to estimate small changes in altitude. Along with direct measurements, air pressure provides a useful indirect indicator of vertical movement. Atmospheric pressure decreases as altitude increases, and this relationship is commonly described using an exponential model. Although accurate, converting pressure measurements into altitude values requires calculations that are too complex to perform repeatedly during...