Salient Object Detection Based on Improved Pyramid Pooling Network
MengHuai Xiao, Yue Wang, Ya Wang, Jun Huang · 2023
Salient object detection is an important task in computer vision, which is to segment visually prominent objects from images accurately. However, image segmentation for small objects and complex background objects is still a challenging task. In order to improve the segmentation effect, a new pyramid-pool network method based on PoolNet network is proposed in this paper. The network, called CDSPNet, integrates Convolutional Block Attention Module (CBAM) and Deep Supervision (DS), wherein the convolutional block attention module can effectively improve the representation ability of feature maps. Depth supervision adds side output to better blend rich deep features. The experimental results show that CDSPNet achieves better salient object detection performance by comparing with other 8 models in F-measure and MAE on 6 public datasets.