Figure A2. The visual comparison of the segmentation results of COVID-19 lung infection compared with other advanced models.

Here (A–C) illustrate three test images obtained from (MedSeg, 2020). For each row, (D) denotes the ground truth, and (E–K) illustrate the segmentation results from FCN8s (Long, Shelhamer & Darrell, 2015), UNET (Ronneberger, Fischer & Brox, 2015), Segnet (Badrinarayanan, Kendall & Cipolla, 2017), Squeeze UNET (Iandola et al., 2016), Residual UNET (Alom et al., 2018), RAD UNET (Zhuang et al., 2019a), ADID-UNET, respectively.