Deep Hashing Image Retrieval Based on Attention Mechanism

In the image nearest neighbor search solution, the deep learning of hashing research as an effective data processing method has made remarkable achievements, especially in large-scale multimedia search applications. However, with the deepening of the convolutional neural network layers, a significant drawback emerges: the loss of important image features. This loss weakens the effectiveness of hash coding and complicates the assessment of high-dimensional data similarity. To overcome these issues, this paper proposes a novel Attention Deep Hashing (ADH) network model. In addition, the Attention mechanism is introduced to extract the image feature encoding weight value for the final encoding results based on the public data set CIFAR-10 datasets. The experimental results demonstrate that the Attention Deep Hash model can effectively improve the MAP value of deep hashing coding and make it more applicable.