Skip to main content

Table 7 Ablation study of P-TransUNet on the Kvasir-SEG dataset for each column

From: P-TransUNet: an improved parallel network for medical image segmentation

Method

mDice

mIoU

Recall

Precision

U-Net [5]

0.5621

0.405

0.4364

0.8466

U-Net++  [6]

0.6783

0.5494

0.7311

0.6885

HRNet-Smallv2 [40]

0.2107

0.1363

0.2038

0.3347

HRNet [40]

0.2349

0.2461

0.3372

0.1523

Deeplabv3 + Xception [37]

0.6746

0.5327

0.6296

0.7757

Deeplabv3 + Mobile [37]

0.6474

0.5098

0.6632

0.6878

MSRF-Net [35]

0.7575

0.6337

0.7197

0.8414

Our method

0.7911

0.6876

0.8409

0.7825

  1. The best results are highlighted. (T means transformer)