Unsupervised hash method based on auto-encoder

Abstract

An unsupervised hash method based on an auto-encoder comprises the steps: establishing a stack type denoising auto-encoder and training the stack type denoising auto-encoder by using a training set, wherein the stack type denoising auto-encoder comprises M encoding layers, a hash layer and M decoding layers which are fully connected in sequence; establishing a stack type auto-encoder with the samestructure as the stack type denoising auto-encoder by using parameters in the stack type denoising auto-encoder, and inputting the images in the training set into the stack type denoising auto-encoder in batches for training; removing a decoder in the final stack-type auto-encoder, and taking the reserved M encoding layers and hash layers as a retrieval network; inputting a to-be-retrieved imageinto the retrieval network to obtain the output of the hash layer; and quantifying an output result of the hash layer to obtain a hash code, calculating a Hamming distance between the to-be-retrievedimage and the hash code of the image in the database, and taking the image with the smallest Hamming distance from the to-be-retrieved image in the database as a retrieval result of the to-be-retrieved image. The unsupervised hash method has excellent retrieval and clustering effects at the same time.

Description

technical field [0001] The invention relates to an unsupervised hashing method based on an autoencoder. Background technique [0002] Hashing, a method of converting high-dimensional feature vectors into binary codes using a mapping function, has achieved notable success in quickly retrieving data. In recent years, the rapid development of convolutional neural network (CNN) has promoted the development of approximate nearest neighbor retrieval. In particular, unsupervised hashing methods have received increasing attention since they do not require labeled training data compared to supervised hashing methods. Restricted Boltzmann machines were first used to encode hash codes in unsupervised hashing methods. However, RBM is basically difficult to implement due to its complexity and the need for pre-training. In recent years, with the development of deep neural networks, especially the development of generative adversarial networks, many studies have achieved remarkable resu...

AI Technical Summary

Problems solved by technology

However, since this retrieval method adaptively allocates different bits to encode data, this method is not an unsupervised learning method, and the encoding result is greatly affected, so further improvement is needed

Images