基于深度学习的手机照片分割  

Mobile phone photo segmentation based on deep learning

作　　者：孙煜 SUN Yu(Qualcomm Enterprise Management(Shanghai)Co.,Ltd.,Shanghai 200000,China)

机构地区：[1]高通企业管理(上海)有限公司,上海20000

出　　处：《计算机应用文摘》2022年第16期92-94,共3页Chinese Journal of Computer Application

摘　　要：随着智能手机的广泛应用,手机照片的人像抠图、人脸美颜、人物特效等应用均依赖照片分割技术。应用基于人工智能的深度学习技术,可以对手机照片进行快速、精确分割。由于智能手机算力有限,文章设计了一种轻量级的深度学习模型,将手机照片进行区域划分,并标注区域属性,提取区域轮廓,可实现快速的人体分割、物体分割、场景分割等功能。同时,在智能手机平台进行照片分割实验,实验结果验证了该方法可行。With the widespread use of smart phones,applications such as portrait matting,face beautification,and character special effects of mobile phone photos all rely on photo segmentation technology.The application of deep learning technology based on artificial intelligence can quickly and accurately segment mobile phone photos.Due to the limited computing power of smartphones,this.paper designs a lightweight deep learning model that divides mobile phone photos into regions,annotates regional attributes,and extracts regional contours,which can achieve fast human body segmentation,object segmentation,scene segmentation and other functions.At the same time,the photo segmentation experiment is carried out on the smartphone platform,and the ex perimental results verify the feasibility of the met hod.

关 键 词：深度学习 分割技术 卷积网络 智能手机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]