Deep hash image retrieval method based on anti-autoencoder
Yu Yang, Yang Liu, Shuang Wang · 2021
Although the supervised deep hash image retrieval method has achieved good performance, its hash function learning process relies on the annotation information of the data set. In reality, most data is very lack of annotation information, so unsupervised deep hash The method can solve this problem well; an unsupervised deep image hash retrieval method based on the anti-self-encoder is proposed. The model is composed of an encoder, a discriminator and a decoder, and a continuous output part is introduced for the encoder to compensate for the hash layer band. For the information loss, the discriminator is used to make the generated hash code more compact. Experiments on single-label dataset MNIST, CIFAR-10, COCO show that the method in this paper has achieved better retrieval performance improvement.