Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules

• DOI: 10.1049/ipr2.12813
• Publication Date: 2023-04-21T01:49:15Z
• Authors: Qingqing Guo; Xianyong Fang; Kaibing Wang; Yuqing Shi; Linbo Wang; Enming Zhang; Zhengyi Liu
• Keywords: Feature (linguistics); Market Segmentation

Abstract The large variations of polyp sizes and shapes the close resemblances polyps to their surroundings call for features with long‐range information in rich scales strong discrimination. This article proposes two parallel structured modules building those features. One is Transformer Inception module (TI) which applies Transformers different reception fields input thus enriches them more scales. other Local‐Detail Augmentation (LDA) spatial channel attentions each block locally augments from complementary dimensions object details. Integrating TI LDA, a new encoder based framework, Parallel‐Enhanced Network (PENet), proposed, where LDA specifically adopted twice coarse‐to‐fine way accurate prediction. PENet efficient segmenting without interference background tissues. Experimental comparisons state‐of‐the‐arts methods show its merits.