Deep Hash Image Retrieval Method Integrating Attention Mechanism
Traditional deep hash-based image retrieval methods focus on some redundant information when obtaining feature information of images,which affects the final image retrieval accuracy.In response to the above issues,this article proposes a fusion cross dimensional interactive attention mechanism module,which can be applied to convolutional neural networks to improve net-work performance and learn more feature information that is conducive to image retrieval.In the deep hash image retrieval task,two classic models,VGG16 and ResNet18,were selected as the basic models for image retrieval.After adding an attention module and redesigning the hash code target loss function,comparative experiments were conducted on the CIFAR-10 and NUS-WIDE datasets.The experimental results showed that the addition of attention mechanism significantly improved the accuracy of image re-trieval,verifying the effectiveness of the proposed method.
万方数据
维普
