图像语义分割深度学习模型综述  

Survey of the deep learning models for image semantic segmentation

作　　者：张新明 祝晓斌[1] 蔡强[1] 刘新亮[1] 邵玮 王磊[3] 

机构地区：[1]北京工商大学计算机与信息工程学院,北京100048 [2]北京中电高科技电视发展有限公司,北京100036 [3]广播科学研究院信息研究所,北京100866

出　　处：《高技术通讯》2017年第9期808-815,共8页Chinese High Technology Letters

基　　金：国家自然科学基金(61402023);北京市自然科学基金(4162019);食品安全知识图谱及大数据平台研制(Z161100001616004)资助项目

摘　　要：阐述了图像语义分割实质上是像素级别的密集分类的概念,及其在计算机视觉领域中的核心地位和应用意义。全面综述了图像语义分割算法的常用分类方法及最新成果,详尽比较了图像语义分割深度学习模型在PASCAL VOC 2012数据集上的像素准确率、平均像素正确率、平均交叠率、频率加权交叠率四个方面的实际表现性能,同时给出了模型的平均耗时、底层框架、实现语言、代码可读性、部署难度信息。最后对语义分割领域的发展进行了总结和展望,指出了模型面临的训练数据集不足、参数优化困难和结构单一的问题。The concept that image semantic segmentation is essentially the dense classification in pixel level,as well as its core position and practical significance are interpreted. Then,the commonly used classifications and latest achievements in image semantic segmentation are comprehensively reviewed,and for some deep learning models for image semantic segmentation,the pixel accuracy,average pixel accuracy,mean intersection over union and frequency-weighted intersection over union on the PASCAL VOC 2012 dataset are compared in detail. Meanwhile,the models＇ other performance indexes including the average time-consuming,program framework,language used,code readability,difficulty of deployment are shown. Finally,the developments of image semantic segmentation are summarized and discussed,and the challeges facing the models such as lack of training data sets,difficulty of parameter optimization and single structure are pointed out.

关 键 词：深度学习 语义分割 PASCAL VOC 卷积神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]