摘要
具备前端图像智能识别能力的终端在电力领域应用广泛,其准确率、漏报率、误报率、效率和功耗等参数与现场应用效果密切相关。为全面、快速的验证智能终端图像识别模型的性能,提出了智能终端图像识别模型测评装置的研制方法,介绍了测评装置的总体架构、硬件组成和软件系统,详细阐述了测试流程,并对多台智能终端进行了测试,与现有模型的提取方法、样本导入方法相比,该装置可以提高智能终端图像识别模型的测试效率,全面评估模型及其所在终端的整体性能。
Terminals with image intelligent recognition capabilities are widely used in the power field,and their performances such as accurate detection rate,missing detection rate,false detection rate,efficiency and power consumption are closely related to application effects.In order to comprehensively and quickly verify the performance of the intelligent image recognition model,the development method of intelligent terminal image recognition model evaluation device has been proposed.Besides,the overall structure,hardware composition and software system are introduced.The detailed test process is explained,and then multiple intelligent terminals have been tested.Compared with the extracting model method and importing sample method,the device can improve the test efficiency and comprehensively evaluate the overall performance of the model and its terminal.
作者
姬艳鹏
陈泽
刘良帅
赵百捷
冯海燕
JI Yanpeng;CHEN Ze;LIU Liangshuai;ZHAO Baijie;FENG Haiyan(State Grid Hebei Electric Power Research Institute,Shijiazhuang 050011,China)
出处
《河北电力技术》
2021年第6期12-15,58,共5页
Hebei Electric Power
关键词
智能终端
图像识别
模型测评
功耗
输电
目标检测
总体架构
intelligent terminal
image recognition
model evaluation
power consumption
power transmission
target detection
overall structure