摘要
目前垃圾分类主要依靠人工来进行,存在效率低、对人体健康有害的问题,文章提出了基于YOLO V3的垃圾自动定位及分类方法。首先,采集公共场所的废弃物图像并进行标注;其次,通过K-mean++确定先验框大小;最后,加载ImageNet数据集上预训练的权重进行迁移训练。结果显示:该方法能够有效完成垃圾的定位及分类,mAP可达82.87%。
In view of the problem of low efficiency and harmful to human health that garbage classification mainly relies on manual work, this paper proposed an automatic garbage location and classification method based on YOLO V3. Firstly, waste images of public places are collected and labeled;secondly, the size of the priors anchor are determined by K-mean++;finally, the weight of pre-training on ImageNet data sets is loaded for migration training. The results show that the method can effectively locate and classify garbage, and the mAP can reach 82.87%.
作者
王铭杰
Wang Mingjie(College of Information Science and Engineering,Shanxi Agricultural University,Taigu 030801,China)
出处
《无线互联科技》
2019年第20期110-112,共3页
Wireless Internet Technology
