多尺度波束注意力卷积网络用于面部表情识别
Jing-Wei Liu1,2, Xiao-Yuan Lin1, Peng-Fei Ji3
1Department of Computer Science, Capital University of Economics and Business, Beijing, 100070, China.
Scientific reports
|July 2, 2025
概括
这项研究通过在卷积神经网络 (CNN) 中引入多级卷积 (MsC) 层和波纹频道注意力 (wCA) 机制来增强面部表情识别,实现了显著的准确性改进.
科学领域:
- 计算机视觉 计算机视觉
- 人工智能的人工智能
- 机器学习 机器学习
背景情况:
- 面部表情识别 (FER) 对人机交互至关重要.
- 目前的卷积神经网络 (CNN) 需要对稳健的FER应用程序进行精度增强.
研究的目的:
- 为了提高面部表情识别系统的准确性.
- 引入新的深度学习架构,以增强FER.
主要方法:
- 通过将第一个卷积层替换为一个多尺度卷积层 (MsC) 提出了多尺度CNN (MCNN).
- 通过结合波形频道注意力 (wCA) 机制引入波形频道注意力CNN (wCA-CNN).
- 开发了基于 wCA 的多尺度 CNN (wCA-MCNN),结合了 MsC 和 wCA.
- 将这些方法应用于基准剩余网络 (ResNet18).
主要成果:
- 与CNN相比,MCNN的准确性提高了1.339%.
- wCA-CNN比CNN提高了1.414%的准确性. wCA-CNN比CNN提高了1.414%的准确性. wCA-CNN比CNN提高了1.414%的准确性.
- wCA-MCNN比CNN获得了2.921%的准确性改进.
- 在ResNet18变种中,改善率高达1.810%.
结论:
- 拟议的MCNN,wCA-CNN和wCA-MCNN架构显著提高了面部表情识别的准确性.
- MsC层和wCA机制的整合为推进FER系统提供了一个有希望的方向.
- 这些方法在现实世界 (FESR) 和标准 (KDEF) 数据集上得到了验证.
相关概念视频
01:21Association Areas of the Cortex
6.3K
Association areas are regions of the cerebral cortex that do not have a specific sensory or motor function. Instead, they integrate and interpret information from various sources to enable higher cognitive processes such as memory, learning, and decision-making. Some key association areas include the following:
Prefrontal Association Area: This area is located in the frontal lobe and is involved in planning, decision-making, and moderating social behavior. It connects with primary motor areas,...
Prefrontal Association Area: This area is located in the frontal lobe and is involved in planning, decision-making, and moderating social behavior. It connects with primary motor areas,...
6.3K
01:24Convolution: Math, Graphics, and Discrete Signals
432
In any LTI (Linear Time-Invariant) system, the convolution of two signals is denoted using a convolution operator, assuming all initial conditions are zero. The convolution integral can be divided into two parts: the zero-input or natural response and the zero-state or forced response, with t0 indicating the initial time.
To simplify the convolution integral, it is assumed that both the input signal and impulse response are zero for negative time values. The graphical convolution process...
To simplify the convolution integral, it is assumed that both the input signal and impulse response are zero for negative time values. The graphical convolution process...
432
01:14Muscles for Facial Expressions
2.8K
The craniofacial muscles are a collection of approximately 20 thin skeletal muscles situated beneath the skin of the face and scalp. These muscles, primarily responsible for the vast array of human facial expressions, originate from the bones or fibrous structures of the skull and extend outwards to connect with the skin. While most skeletal muscles in the body are enveloped in thick fascia, facial muscles generally have a more delicate fascial covering, with the buccinator muscle being a...
2.8K
01:17Convolution Properties II
292
The important convolution properties include width, area, differentiation, and integration properties.
The width property indicates that if the durations of input signals are T1 and T2, then the width of the output response equals the sum of both durations, irrespective of the shapes of the two functions. For instance, convolving two rectangular pulses with durations of 2 seconds and 1 second results in a function with a width of 3 seconds.
The area property asserts that the area under the...
The width property indicates that if the durations of input signals are T1 and T2, then the width of the output response equals the sum of both durations, irrespective of the shapes of the two functions. For instance, convolving two rectangular pulses with durations of 2 seconds and 1 second results in a function with a width of 3 seconds.
The area property asserts that the area under the...
292