An Image Retrieval Method Based on Ranking Learning and Multivariate Loss

Abstract

The invention discloses an image retrieval method based on ranking learning and multivariate loss. The core idea of ​​the method is to obtain the sequence number in the similarity ranking of the query image while selecting the negative sample group for the query image, and sort the The sequence number and features are combined to obtain the loss function and update the network, so as to accurately extract image features. The invention introduces the theory of ranking learning into image retrieval, adjusts network parameters according to the Euclidean distance between negative samples and query pictures, and can learn image features more comprehensively to perform more accurate retrieval. The present invention fully considers the impact of negative samples on experiments, and can adjust the number of negative samples according to the training effect of the model.

Description

An Image Retrieval Method Based on Ranking Learning and Multivariate Loss technical field The present invention belongs to the technical field of image retrieval, relate to a kind of image retrieval method based on sorting learning and multivariate loss Law. Background technique [0002] With the large-scale popularization of digital cameras and smart phones in recent years and the continuous increase in the capacity of storage devices, many Media content, especially visual data, has shown an explosive growth trend. For massive visual content, how to quickly Effective retrieval is a research hotspot in academic and industrial circles at home and abroad. [0003] At present, the method of large-scale search engines for Internet image retrieval is mainly based on the pre-trained classification network. The network is initialized and trained for different tasks, which is called a fine-tuning network. Validation-based fine-tuning The network is mainly used in image ret...

AI Technical Summary

Problems solved by technology

The model trained by ternary loss has a lot of randomness when selecting samples, which takes a long time, which will lead to relatively large intra-class distance, and has weak generalization ability from training to testing.

Thus quadruple networks, hard sampled triplets, and boundary sample mining networks emerged, however, these Siamese networks usually rely on simpler network architectures than the one we use here, which involves several regions The collection and aggregation of image retrieval has low accuracy and robustness. More importantly, the existing metric learning network performs feature learning by pulling the positive samples closer and pushing away the negative samples. In the distance setting, the same value is used, but not all negative samples have the same dissimilarity with the query image, so this design is difficult to accurately extract the features of the image

Images