Deep Hash has been applied in the field of image search very well. However, the previous deep Hash method has the limitation that the semantic information is not fully utilized. This paper, develops a discrete Hash algorithm based on deep supervision, assuming that learning binary code should be an ideal choice of classification. The pair tag information and classified information are used to learn Hash codes within a framework. The output of the last layer is restricted to binary code directly. Due to the discrete properties of Hash codes, the alternate minimization method is used to optimize the target function. The proposed algorithm is proved to be better than the other supervised Hash methods in three image retrieval databases CIFAR-10,NUS-WIDE and SUN397.%深度哈希在图像搜索领域取得了很好的应用,然而,先前的深度哈希方法存在语义信息未被充分利用的局限性.开发了一个基于深度监督的离散哈希算法,假设学习的二进制代码应该是分类的理想选择,成对标签信息和分类信息在一个框架内用于学习哈希码,将最后一层的输出直接限制为二进制代码.由于哈希码的离散性质,使用交替最小化方法来优化目标函数.该算法在三个图像检索数据库CIFAR-10、NUS-WIDE和SUN397中进行验证,其准确率优于其他监督哈希方法.
展开▼
