Lightweight Region Extraction in Small Animal Images Based on Salient Object Detection

This paper proposes an unsupervised salient objects detection technique tailored for objects extraction in small animals, addressing the limitations of existing methods in terms of model complexity and small object detection capabilities. By introducing a background generation module and a lightweight model, it enhances the understanding of image edge information and semantic information of salient objects. Experiments demonstrate that this method outperforms existing techniques under unsupervised conditions and exhibits excellent performance in object extraction for small animal categories.