摘要
本系统通过在嵌入式平台运行神经网络框架TensorFlow为基础来构建卷积神经网络,用手写数字数据集MNIST来训练构建好的神经网络。由于嵌入式平台的计算能力偏低,此外训练的过程会消耗大量的资源,则可以利用PC机资源训练已设计好的网络、参数,最后将训练好的参数加载到嵌入式平台。ARM处理器因为其优异的性能和极低的功耗以及丰富而强大的功能而在市场有着很高的占有率,具有接口方便,便于软件开发的特点。
Based on running the neural network framework Tensor Flow on an embedded platform, this system builds the convolutional neural network, which is trained by MNIST handwritten digital data. Due to the low capacity of the embedded platform, in addition, the process of training will consume large amounts of resources, the PC resource can be used to train the network and parameter that already designed, finally the trained parameters are loaded into the embedded platform. Arm processors have a high market share because of their excellent pefformance, extremely low power consumption and rich and powerful functions. It also has featmres of convenient interface and easy for softwmre development
作者
杨迪
黄盼盼
夏志勇
沈森
Yang Di, Huang Panpan,Xia Zhiyong, Shen Sen,(School of Inormation Engineering, Henan University of Science and Technology, Luoyang Henan 471003, China)
出处
《山西电子技术》
2018年第5期22-24,34,共4页
Shanxi Electronic Technology
基金
河南科技大学SRTP(2017075)资助课题
关键词
手写数字识别
卷积神经网络
ARM系统
handwritten numeral recognition
convolution neural network
ARM system
