Deep learning of Hash Method for Fast Image Retrieval

Most of the image retrieval technology are complex and advanced methods to improve the accuracy of retrieval.They also have deep net layers, which costs much time to caculate the result. When we need deal with complex image task, it will be hard to complete the fast image retrieval. This paper presents a convolutional neural network combining with the fast encoding technology under the hash algorithm to improve fast retrieval time and accuracy. In order to keep the existing accuracy and retrieval speed of the network, we present a Hash method path of the network combining witn the dual networks.By clustering analysis and AlexNet,we get lots of high dimensional feature. Then these feature is put into Hash-coding algorithm.The algorithm will output binary feature to match the target image. Through learning similar graph pairs, Hash net work shorten the distance between the similar images, and achieve fast clustering.Then it realizes image retrieval and outputs the most similar picture.Complete the search task.