Dataset-level color augmentation and multi-scale exploration methods for polyp segmentation

Open Webpage

Haipeng Chen, Honghong Ju, Jun Qin, Jincai Song, Yingda Lyu, Xianzhu Liu

Published: 01 Jan 2025, Last Modified: 18 Nov 2024Expert Syst. Appl. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We address the color appearance imbalance in limited polyp datasets.•Dataset-level color augmentation (DLCA) decouples the color and content of images.•DLCA innovatively uses dataset-level color statics knowledge for the first time.•We achieve outstanding results on ETIS, reaching mDice of 0.839 and mIoU of 0.766.•The proposed approaches are universal, practical, and convenient.