Mobile terminal-oriented instance-level image retrieval method and device

Abstract

The invention discloses a mobile terminal-oriented instance level image retrieval method and device, and the method comprises the steps: constructing and training a lightweight neural network classification model which comprises an embedded layer, a convolution attention module, seven bottleneck blocks, a full connection layer, a convolution attention module, a binary adaptive mean pooling layer and a classifier; and removing the last layer of classifier of the lightweight neural network classification model to serve as a feature extractor, extracting the image features of the to-be-retrieved image by adopting the feature extractor, calculating the Euclidean distance between the image features of the to-be-retrieved image and the image features of the retrieval data set image, and outputting the retrieval data set image corresponding to the minimum Euclidean distance. According to the method, the problem that a deep neural network model is difficult to deploy in mobile equipment in deep learning is solved, and the lightweight neural network is successfully realized in an image retrieval task at a time.

Description

technical field [0001] The present application belongs to the technical field of image retrieval, and in particular relates to an instance-level image retrieval method and device for mobile terminals. Background technique [0002] Content-based image retrieval, by extracting the visual information of images, can quickly and accurately retrieve the images needed by users from a large number of digital images. This research has great value both in the research field and in commercial applications. In recent years, with the rapid development of deep learning methods, thanks to the accurate expression of image content by deep features, significant progress has been made in using deep learning models to retrieve images. [0003] However, the computation and parameters of most convolutional neural networks are large, and most state-of-the-art convolutional neural networks are difficult to deploy on resource-constrained mobile devices due to the limitations of storage space and pow...

AI Technical Summary

Problems solved by technology

Therefore, many typical lightweight neural networks have been proposed in the industry, such as MobileNetV2. Although the network reduces parameters and calculations, it still has a lot of redundancy.

Images