目标多种多值属性的端端快速识别网络  

End to End Object Recognition Algorithm for Multi-attributes of Multi-values

作　　者：周伦钢 孙怡峰[2] 王坤 吴疆 黄维贵 李炳龙[2] ZHOU Lungang;SUN Yifeng;WANG Kun;WU Jiang;HUANG Weigui;LI Binglong(Henan Industrial School,Zhengzhou 450002,China;Information Engineering University,Zhengzhou 450001,China;Zhengzhou Xinda Institute of Advanced Technology,Zhengzhou 450001,China)

机构地区：[1]河南省工业学校,郑州450002 [2]信息工程大学,郑州450001 [3]郑州信大先进技术研究院,郑州450001

出　　处：《计算机工程与应用》2021年第9期182-190,共9页Computer Engineering and Applications

基　　金：郑州市重大科技创新专项项目(188PCXZX773);工信部2018年大数据产业发展试点示范项目。

摘　　要：为提高图像目标多种多值属性的识别速度,提出一种端到端的识别算法。采用修正的YoloV3网络作为主网络,确定目标的boundingbox;依据属性独立特性构造子网络,多个子网络共享由boundingbox确定的主网络深层次特征,进行推断,并采用多值输出满足多值属性的识别。在训练过程中,采用了三阶段分目标训练。实验结果验证了该算法在识别准确度和时间效率上的优良性能。In order to improve image object recognition speed for multi-attributes of multi-values,an end-to-end recognition algorithm is proposed.Firstly,the modified YoloV3 network is used as main network in order to detect the object bounding boxes.Sub-networks are constructed according to the independent attributes.Sub-networks share the deep bounding box features of main network and adopt multi-outputs to recognize the attributes multi-values.There are three stages with different objective functions in the training process.Experimental results show that the proposed algorithm has good performance.

关 键 词：目标检测 属性识别 深度学习 卷积神经网络 图像识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]