摘要
提出了一种新的基于人脸表情识别的图像理解模型。由于传统的图像理解模型更多地关注图像中事物之间的联系而忽略事物的具体细节,对图像的理解不够充分。针对以人为主体的图像,在提取图像特征时加入一个表情识别卷积神经网络,并改进了各层参数。实验证明,与传统方法相比,改进的图像理解模型能够识别出图像中人物的情绪,对图像的理解更加清晰准确。
This paper presents an improved image understanding model based on facial expression recognition.Generally speaking,an image understanding model mainly focuses on the relation between objects in the image.As a result,the features of one specific object can't be detected.Yet understanding the inner emotion from one image is a pressing need in many areas.To solve this problem,it extendes the traditional model by adding expression recognition process during extracting image features and improved the parameters of CNN and LSTM.
出处
《工业控制计算机》
2018年第3期34-36,共3页
Industrial Control Computer
基金
上海市科学技术委员会科研计划项目(17511106802)
关键词
图像理解
表情识别
深度学习
image understanding
expression recognition
deep learning