摘要
为解决当前流行的哈希检索方法生成的哈希码存在信息冗余,不能很好地保留图像语义相似性等问题,提出一种基于深度卷积神经网络来学习二进制哈希编码的方法。利用深度卷积神经网络提取图像的特征表示;将来自两个完全连接层的图像特征表示输入到哈希层,将分类误差以及阈值误差添加到损失函数中进行训练;将查询图像输入模型得到对应的哈希码。在CIFAR-10和NUS-WIDE两个数据集上进行实验,实验结果表明,所提方法在检索精度方面优于其它现有哈希方法。
To solve the problem that the current popular hashing retrieval method can not keep the semantic similarity of images well,and the generated hashing codes have information redundancy and so on,a method based on deep convolutional neural network to learn binary hashing coding was proposed.The deep convolutional neural network was used to extract the feature representation of the image.The image feature representation from two fully connected layers was entered into the hash layer,and the classification error and threshold error were added to the loss function for training.The query image was entered into the model to obtain the corresponding hash codes.Results of experiments on CIFAR-10 and NUS-WIDE show that the proposed method is superior to other existing hash methods in terms of retrieval accuracy.
作者
冯兴杰
程毅玮
FENG Xing-jie;CHENG Yi-wei(School of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China)
出处
《计算机工程与设计》
北大核心
2020年第3期670-675,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(U1233113)
国家自然科学青年基金项目(61301245、61201414)。
关键词
图像检索
哈希
深度卷积神经网络
信息冗余
均值平均精度
image retrieval
hash
deep convolutional neural network
information redundancy
mean average precision
