 This paper proposes a novel evaluation method for selecting the most suitable convolutional neural network, CNN, for automatic gastric polyp segmentation. It compares 7 different CNN models, including UNET, UNET++, D-Lab 3, D-Lab 3+, PAN, LinkNet, and MANet, using a combination of subjective and objective measures. The results show that UNET++ with the MobileNet V2 encoder achieves the highest accuracy and is therefore selected to construct the automated polyp segmentation system. This study demonstrates the potential of CNNs for the diagnosis of gastric polyps and provides a useful tool for selecting the most appropriate CNN model for clinical applications. This article is authored by Taoyan, Yingqin, Pat Kinwong, and others.