Semantic segmentation of small intestinal polyps based on dual-branch network
Intestinal polyps are recognized as the primary precursor lesions of colorectal cancer,and their early detection and accurate segmentation are considered of great significance for cancer prevention.To address over-reliance on a large number of pixel-by-pixel annotated samples and poor generalization to unknown polyp regions.A few-shot intestinal polyp image semantic seg-mentation method based on a dual-branch network is proposed.A dual-branch network with support and query branches is estab-lished,and the interactive information between the support and query branches is used to guide mask prediction of unknown re-gions in the query branch.Tests are conducted on multiple intestinal polyp image datasets,and the results indicate that the pro-posed method achieves higher segmentation performance.
万方数据
