Method for detecting image salient target

Abstract

The invention provides a method for detecting an image saliency target, and belongs to the technical field of image saliency detection. The method comprises the following steps: in a first stage, constructing a classification network CNet and a character sequence generation network PNet to obtain saliency maps from classification and character sequence generation tasks respectively; in a second stage, establishing two complementary training data sets, namely a natural image data set with a noise label and a network image data set, by utilizing CNet and PNet, and alternately training SNet by utilizing the two complementary data sets; and in the third stage, updating the natural image data set and the network image data set by using the prediction result of the SNet, and recursively optimizing the model. In the test stage, only the SNet is used to predict the saliency maps. Experiments show that the method is superior to unsupervised and weakly supervised methods, and still has the goodperformance compared with some supervised methods.

Description

technical field [0001] The invention belongs to the technical field of image saliency detection, and aims to detect a salient object in any image, so as to segment the most salient object area in the image. Background technique [0002] Image saliency detection has attracted widespread attention in recent years. As a preprocessing method, image saliency detection has been widely used in many fields such as image compression, image classification, and image segmentation. Early saliency detection research mainly used manually designed features and heuristic priors to predict salient regions in images, such as using center priors, background priors, etc. In recent years, with the successful application of deep convolutional neural networks (CNNs) in various vision tasks, many deep learning-based saliency detection methods have been proposed. In 2015, the paper "Visual saliency based on multi-scale deep features" published in CVPR proposed to extract multi-scale features from d...

AI Technical Summary

Problems solved by technology

Consequently, the results of training saliency models using only image-level labels often lack sharp boundaries between object and background regions, or misclassify pixels near salient object contours.

Images