Applied Sciences, Год журнала: 2024, Номер 15(1), С. 173 - 173

Опубликована: Дек. 28, 2024

Camouflaged object segmentation refers to the task of accurately extracting objects that are seamlessly integrated within their surrounding environment. Existing deep-learning methods frequently encounter challenges in segmenting camouflaged objects, particularly capturing complete and intricate details. To this end, we propose a novel method based on Curiosity-Driven network, which is motivated by innate human tendency for curiosity when encountering ambiguous regions subsequent drive explore observe objects’ Specifically, proposed fusion bridge module aims exploit model’s inherent fuse these features extracted dual-branch feature encoder capture details object. Then, drawing inspiration from curiosity, curiosity-refinement progressively refine initial predictions exploring unknown object’s Notably, develop curiosity-calculation operation discover remove leading accurate results. Extensive quantitative qualitative experiments demonstrate model significantly outperforms existing competitors three challenging benchmark datasets. Compared with recently state-of-the-art method, our achieves performance gains 1.80% average Sα. Moreover, can be extended polyp industrial defects tasks, validating its robustness effectiveness.

Язык: Английский