Cross-modal hash retrieval method based on long and short mapping

The invention relates to the technical field of cross-modal retrieval, and discloses a cross-modal Hash retrieval method based on long and short mapping, which comprises the following steps: establishing a multi-modal data set which comprises a plurality of image text description pairs consisting of images and corresponding text descriptions; constructing a long-short mapping method; constructing a retrieval model; performing hash retrieval; according to the method, the multi-dimensional features are extracted through deep learning, and mapping among different-dimensional features is realized according to the correlation difference among the multi-dimensional features, so that the expression ability of the low-dimensional features is enhanced, and the retrieval accuracy is improved. 本发明涉及跨模态检索技术领域，公开了一种基于长短映射的跨模态哈希检索方法，包括以下步骤：建立多模态数据集，多模态数据集中包括多个由图像以及对应的文本描述组成的图像文本描述对；构建长短映射方法；构建检索模型；哈希检索；本发明通过深度学习提取多维度特征，依据多维度特征之间的相关性差异实现不同维度特征之间的映射以增强低维特征的表现能力并提升检索准确率。