摘要
TensorFlow是Google基于DistBelief研发的第二代人工智能学习系统,文章在此平台上基于MobileNet模型,利用超市中商品的图片进行迁移学习,最终获得五个对超市商品有较高识别率的模型,最高测试集精度可达87.4%,具有模型小、训练时间短的优点,可部署于无人超市及物流分拣等领域。
TensorFlow is the second generation artificial intelligence learning system based on DistBelief developed by Google.The research in this paper is based on MobileNet model on this platform. By using the pictures of goods in the supermarket to carry on transfer learning, fivemodels withhigh recognition rate to the goodshave been obtained, which has the advantages of small model and short training time, and the highest test set accuracy is up to 87.4%. Besides, this model can be easily deployed in several areas such as unmanned supermarkets and logistics sorting.
作者
王傲然
刘玮
WangAoran;Liu Wei(School of Infonnation and Electrical Engineering,Zhejiang University City College,Hangzhou 310015,China)
出处
《信息通信》
2018年第12期49-50,共2页
Information & Communications
基金
浙江大学城市学院大学生科研训练项目(XZ2018522038)
浙江省高等教育学会2014年度高等教育研究课题
杭州市科技局规划项目(20160533B96)
