摘要
细粒度图像识别是计算机视觉领域的一个难题,也是大数据时代和人工智能的一个具有挑战性的难题。为了解决该问题,提出了一种基于双线性卷积网络的对象定位方法,该方法通过双线性网络来得到VGG网络不同层的特征,将其交集作为最终的定位框。为了验证所提方法的可行性,在CUB-200-2011数据集和Stanford Cars-196上对所提方法进行验证,实验结果表明所提的方法的对象定位精确度远远高于其他具有代表性的方法,同时在加入了软最大化分类器后,所提方法得到的分类准确率比原始图像有了较大提高。
Fine-grained Image recognition is a difficult problem for computer vision area,and it is also a challenge for big data and artificial intelligence.In order to solve this problem,an object localization method based on bilinear convolutional network is proposed.The bilinear convolutional network is used to obtain different features of VGG network,and the intersection set is used as the localization framework.In order to verify the feasibility of this paper,the proposed method is simulated on CUB-200-2011 and Stanford Cars-196.The simulation result shows the accuracy of the proposed method is higher than other representative methods.After adding the Softmax classier,this proposed method improved largely compared with the primitive image as the input.
作者
陈斯祺
CHEN Si-qi(First High School of Fu’an City,Fu’an 355000,Fujian)
出处
《电脑与电信》
2018年第11期78-80,共3页
Computer & Telecommunication
关键词
细粒度识别
对象定位
特征
双线性网络
fine-grained recognition
object localization
feature
bilinear network
