Printing roller image salient object detection based on multi-scale feature fusion

Abstract In this study, aiming at the problem that surface defect image of printing roller has different sizes and a large number defects, salient object detection method based on stepwise multi-scale feature fusion is proposed, which consists guidance module, multi-resolution information deeply supervised side output. Specifically, module contains two paths, top-down bottom-up, strengthen transmission features in network, fuse multi-level through to suppress interference background noise redundant features. addition, outputs are used eliminate differences caused by foreground imbalance. The experimental results dataset show can achieve Fmax , MAE 0.7959, 0.0144, rate reach 24.98fps, better performance than other algorithms.