Camouflaged Object Segmentation With A Single

In the field of computer vision, camouflage object segmentation tasks typically require a large amount of annotated data to achieve effective segmentation. However, promptable segmentation models like the Segment Anything Model (SAM) can perform excellently on unseen images with just instance-specific visual prompts. For complex scenes with camouflaged objects, SAM's performance may still be limited even with instance-specific prompts. Therefore, this task aims to improve the performance of camouflage object segmentation across different datasets through a single task-general prompt, thereby reducing the reliance on large amounts of labeled data and enhancing the model's generalization and practicality.