Face Emotion Recognization Using Dataset Augmentation Based on Neural Network

2025-05-06 0 0 838.41KB 5 页 10玖币

侵权投诉

Face Emotion Recognization Using Dataset Augmentation Based

on Neural Network

Mengyu Rao

Fuzhou University

Ruyi Bao

University of Nottingham

Liangshun Dong*

Shanghai Jiao Tong University

ABSTRACT

Face expression plays a critical role during the daily life, and peo-

ple cannot live without face emotion. With the development of

technology, many methods of facial expression recognition have

been proposed. However, from traditional methods to deep learning

methods, few of them pay attention to the hybrid data augmenta-

tion, which can help improve the robustness of models. Therefore,

a method of hybrid data augmentation is highlighted in this paper.

The hybrid data augmentation is a method of combining several

eective data augmentation. In the experiments, the technique is

applied on four basic networks and the results are compared to the

baseline models. After applying this technique, the results show

that four benchmark models have higher performance than those

previously. This approach is simple and robust in terms of data aug-

mentation, which makes it applied in the real world in the future.

Besides the results show versatility of the technique as all of our

experiments get better results.

CCS CONCEPTS

•Computing methodologies →Articial intelligence.

KEYWORDS

Deep learning, Computer vision, Facial expression recognition,

Facial emotion

ACM Reference Format:

Mengyu Rao, Ruyi Bao, and Liangshun Dong*. 2022. Face Emotion Recog-

nization Using Dataset Augmentation Based on Neural Network. In 2022

The 6th International Conference on Graphics and Signal Processing (ICGSP)

(ICGSP 2022), July 1–3, 2022, Chiba, Japan. ACM, New York, NY, USA, 5 pages.

https://doi.org/10.1145/3561518.3561519

1 INTRODUCTION

Facial expression is one of the most external indications of a per-

son’s feelings and emotions. In daily conversation, according to

the psychologist, only 7% and 38% of information is communicated

through words and sounds respective, while up to 55% is through

facial expression [

]. It plays an important role in coordinating

interpersonal relationships. Ekman and Friesen [

] recognized six

Permission to make digital or hard copies of all or part of this work for personal or

classroom use is granted without fee provided that copies are not made or distributed

for prot or commercial advantage and that copies bear this notice and the full citation

on the rst page. Copyrights for components of this work owned by others than ACM

must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,

to post on servers or to redistribute to lists, requires prior specic permission and/or a

fee. Request permissions from permissions@acm.org.

ICGSP 2022, July 1–3, 2022, Chiba, Japan

ACM ISBN 978-1-4503-9637-0/22/06. . . $15.00

https://doi.org/10.1145/3561518.3561519

essential emotions in the nineteenth century depending on a cross-

cultural study [

], which indicated that people feel each basic emo-

tion in the same fashion despite culture. As a branch of the eld of

analyzing sentiment [

], facial expression recognition oers broad

application prospects in a variety of domains, including the inter-

action between humans and computers [

], healthcare [

], and

behavior monitoring [

]. Therefore, many researchers have de-

voted themselves to facial expression recognition. In this paper, an

eective hybrid data augmentation method is used. This approach

is operated on two public datasets, and four benchmark models see

some remarkable results.

2 RELATED WORKS

2.1 VggNet

The VGG model [

] was posted by the Visual Geometry Group

team at Oxford University. The primary goal of this architecture

is to demonstrate how the its nal performance can be impacted

by increasing network depth. In VGG, 7

7 convolution kernels are

replaced by three 3

3 convolution kernels, and 5

5 convolution

kernels are replaced by two 3

3 convolution kernels. The main

goal of the change is to make sure that the depth of the network

and the impact of the neural network can be ameliorated with the

condition of the same perceptual eld.

2.2 ResNet

The ResNet [

] model won rst place in the ImageNet competition

[

] held in 2015. The problem that deepening the model can decrease

the accuracy was solved by this work. Due to the proposed residual

block, it is easy to learn the identity mapping, even though stacked.

If there are numerous blocks, redundant blocks can also learn the

identity mapping with the help of the residual block. Furthermore, it

improves the eectiveness of SGD optimization, which can optimize

the network in deeper. What is more, no additional parameters

and computational complexity are introduced. Only a very simple

addition operation is performed and the complexity is negligible

compared to the convolution operation. The ResNet architecture is

shown in Figure 1.

Figure 1: The structure of ResNet

2.3 Xception

The Xception [

] model is an upgraded version of the InceptionV3

[

] model. Chollet F oers a new structure of deep convolutional

neural network named Xception that replaces the Inception module

arXiv:2210.12689v2 [cs.CV] 21 Nov 2022

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

FaceEmotionRecognizationUsingDatasetAugmentationBasedonNeuralNetworkMengyuRaoFuzhouUniversityRuyiBaoUniversityofNottinghamLiangshunDong*ShanghaiJiaoTongUniversityABSTRACTFaceexpressionplaysacriticalroleduringthedailylife,andpeo-plecannotlivewithoutfaceemotion.Withthedevelopmentoftechnology,manymetho...

展开>> 收起<<

Face Emotion Recognization Using Dataset Augmentation Based on Neural Network.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Face Emotion Recognization Using Dataset Augmentation Based on Neural Network

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: