The binary hash code generated by the traditional deep hash image retrieval method has information redundancy and cannot reflect the local semantic information of the image well. This paper proposes a deep hash image retrieval method that combines a convolutional neural network with an attention model. It uses the VGG16 network as the image feature extractor, and then adds an attention module after the convolutional layer of the model to refine the feature map, and finally output the binary hash code as the feature of the image in the fully connected layer of the model, thereby improving the accuracy of the image retrieval task. Experiments on the CIFAR-10 and NUS-WIDE datasets show that after the attention mechanism is added, the model uses different digit binary hash codes in the two datasets to achieve the highest retrieval accuracy of 85.3% and 78.1%, higher than the case where the attention mechanism is not used, verifying the effectiveness of attention model.