Attention mechanism-based Image collaborative salient target detection model

Abstract

The invention provides an attention mechanism-based image collaborative salient target detection model, which is characterized by comprising the following steps: firstly, extracting features of a single image; secondly, extracting channel characteristics of each image by utilizing the channel attention of the single image and the channel collaborative attention of the multiple images; thirdly, extracting the spatial features of each image by using the spatial attention of the single image and the spatial collaborative attention of the multiple images; then, generating a collaborative saliency map through a decoder; and finally, carrying out learning training on the model under the supervision of the significant true value graph. And the model activates common category information of images in the group through channel collaborative attention, and activates positions of similar targets of the images in the group through space collaborative attention. The model utilizes an attention mechanism to mine the collaborative relationship of the images in the group, and the effect of image collaborative salient target detection is improved.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to an image collaborative salient target detection model based on an attention mechanism. Background technique [0002] Finding common salient objects in a group of images is called image co-salient object detection. Traditional methods mainly use hand-painted features, such as color, contrast, and contextual features. For example, in 2013, Fu et al. "Cluster-based co-saliency detection" uses clustering methods to use repeated attributes as additional constraints to discover common features in a group of images. significant object. Existing image collaborative salient target detection methods based on deep learning mainly use convolutional neural networks to mine collaborative relationships. For example, in 2017, Wei et al. "Group-wise deep co-saliency detection" designed a fully convolutional neural network for image collaboration. For salient object detection, the result of cascad...

AI Technical Summary

Problems solved by technology

[0003] In the above methods, hand-painted low-level features rely on prior knowledge, and its accuracy often depends on subjective judgments. Therefore, it is difficult to obtain very good detection results in complex scenes, such as the method of Fu et al. in 2013; and depth The learning method uses simple cascading operations to achieve collaborative feature extraction, which cannot fully mine the relationship and similar features between images, such as the methods of Wei et al. in 2017 and Ren et al. in 2019.

Images