Target saliency detection method based on step-by-step attention mechanism

Abstract

The invention discloses a target saliency detection method based on a step-by-step attention mechanism. The method comprises the following steps: firstly, carrying out image data preprocessing to obtain a preprocessed image training set, then constructing a target saliency detection network, and finally training the target saliency detection network through the image training set; according to the method, the feature information is progressively progressive on the space attention module and the channel attention module, the channel attention module group and the space attention module group enable the network to extract richer context information, semantic mixing is avoided, segmentation noise, especially false negative noise, is effectively reduced, and the segmentation efficiency is improved. And the target saliency detection performance is remarkably improved.

Description

technical field [0001] The invention belongs to the field of computer vision and relates to the fields of salient target detection and image detection. Specifically, it involves an object saliency detection method based on a hierarchical attention mechanism. Background technique [0002] With the rapid development of deep learning and neural networks, the field of computer vision has achieved an unprecedented leap. As a classic category in the field of computer vision, object detection has been extensively researched and discussed, and has made great progress in various directions such as salient object detection, pedestrian re-identification, and image data evaluation. In daily life, face scanning, license plate scanning, Skynet engineering, etc. all use the related technologies of target detection. [0003] The human visual system has the ability to quickly search and locate objects of interest when facing natural scenes. This visual attention mechanism is an important m...

AI Technical Summary

Problems solved by technology

[0006] However, the existing salient target detection network based on the attention mechanism often uses the attention module alone, or simply integrates the attention module into the encoding and decoding process as a priori module to improve its own network performance, and fails to fully tap the attention. The performance improvement and application of the module to the network, the wrong use often leads to the confusion and noise of semantic information

Images