Scholarlink

Facial expression recognition (FER), also known as facial emotion detection, is an emerging domain within computer vision and machine learning. This field has broad implications to education, psychology human–computer interaction and market analysis, etc. Facial expressions recognition need to correctly identify facial expressions in order to solve requests of these areas. In this work, we investigate emotion from facial expression detection in FER-2013 dataset by just extracting label. It investigates the accuracy of two different Convolutional Neural Network (CNN) architectures are: Inception-V3 and Sequential model. So, we have to classify facial expressions (Accounting for 7 classes: anger, fear, disgust, happiness, surprise, sadness, and a neutral category). Empirical results also indicates that the Inception-V3 model fine-tuned performs better than the one most sequential model for predicting the labels of FER-2013.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Detection of Facial Emotion with Deep Learning Models and Contribution of Inception-V3

  • 章节

  • 发表日期：

    2025年

查看全文
  • Fouad Lehlou,
  • Adil El Makrani,
  • Jalal Laassiri

摘要

Facial expression recognition (FER), also known as facial emotion detection, is an emerging domain within computer vision and machine learning. This field has broad implications to education, psychology human–computer interaction and market analysis, etc. Facial expressions recognition need to correctly identify facial expressions in order to solve requests of these areas. In this work, we investigate emotion from facial expression detection in FER-2013 dataset by just extracting label. It investigates the accuracy of two different Convolutional Neural Network (CNN) architectures are: Inception-V3 and Sequential model. So, we have to classify facial expressions (Accounting for 7 classes: anger, fear, disgust, happiness, surprise, sadness, and a neutral category). Empirical results also indicates that the Inception-V3 model fine-tuned performs better than the one most sequential model for predicting the labels of FER-2013.