Hash center-based continuous learning method

Abstract

The invention relates to a hash center-based continuous learning method. The method comprises the steps that similar data pairs and non-similar data pairs in an image data set are input into a convolutional neural network for feature learning; a feature learning result passes through a hash layer, the hash layer comprises three full connection layers, and each full connection layer comprises a rowof neurons and an activation function; the hash codes of the similar data pairs converge to a public hash center after being subjected to center similarity constraint, the hash codes of the non-similar data pairs converge to different hash centers after being subjected to center similarity constraint, and the hash layer converts continuous depth representation into K-dimensional representation; after the hash layer outputs a real number vector, the K-dimensional representation is binarized into a K-bit binary hash code by using an activation function. According to the method, richer image features can be learned, less detail information of data is lost, the generated binary hash code is higher in accuracy and high in distinguishability, and the image hash retrieval performance can be improved.

Description

[0001] (1) Technical field: [0002] The invention relates to a machine learning method, in particular to a continuous learning method based on a hash center. [0003] (two), background technology: [0004] Image retrieval, that is, given a query image, quickly and accurately return similar images from the specified database, and obtain image sequences according to the similarity measure. Image retrieval can be divided into two categories according to different descriptions of image content: text-based image retrieval (Keyword-Based Image Retrieval, KBIR) and content-based image retrieval (Content-Based Image Retrieval, CBIR). [0005] Text-based image retrieval systems rely on keyword expressions to match query images with database images, and then return the results that best match the keyword content. The advantage of this text-based image retrieval system is that the system is relatively simple to implement and the retrieval speed is relatively fast. The disadvantage is th...

AI Technical Summary

Problems solved by technology

[0014] (3) Discrete optimization problems in hash learning

However, the current general methods all have NP-hard (Non-deterministic Polynomialtime Hardness) problems, which cannot effectively solve this discrete optimization problem under the condition of large-scale image data.

[0016] The existing deep hash learning method mainly learns continuous hash representation through the similarity of data pairs, but only through the local information of data pairs cannot effectively learn the global distribution of large-scale data.

At the same time, affected by the ill-posed gradient problem in symbolic activation function optimization, most methods need to learn continuous representations first, and then generate binary hash codes in a separate binarization process, which will seriously lose the original feature details of the data and reduce retrieval performance.

Images