Research on Collaborative Salient Object Detection Using Collaborative Feature Integration and Extraction

Collaborative salient object detection is a critial preprocessing part of many complex computer vision tasks. Therefore, the study of co-salient object detection in complex scenes has important practical significance and value. This research presents a deep learning-based model that utilizes collaborative feature extraction and fusion. First, the salient object detection algorithm is designed to obtain single image salient feature maps(SISMs), combining VGG16 and normalized masking average pooling(NAMP) approach to fetch images. Then the correlation information is obtained by feeding the internal information and category features of a single image into the correlation fusion module. Additionally, to further enhance the performance, the self-correlation characteristics are rearranged. Finally, the proposed model and four advanced algorithms were tested on the public authoritative datasets. The result showed that our method had better performance in co-saliency detection, and the ablation study verified the effectiveness of each module.