Figure 11. The qualitative results of comparative experiments with or without Attention Gate in the network. Here (A–C) are three test images.

For each row, (D) denotes the ground truth, (E–I) illustrate the segmentation results from UNET, AG-UNET, DA-UNET, IDA-UNET and ADID-UNET, respectively. AG-UNET– the addition of AG module to UNET, DA-UNET–adding two dense networks and AG module to the network without including the IDC module. IDA-UNET refers to adding IDC and AG modules to the UNET without adding dense networks, and ADID-UNET indicates that dense networks, IDC and AG module are added to the network.