Skip to main content

Table 5 Generalizability results of the models trained on Kvasir-SEG and tested on CVC-clinicDB

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

Method

mDice

mIoU

Recall

Precision

U-Net [5]

0.6302

0.5015

0.5612

0.8249

U-Net++  [6]

0.4267

0.3623

0.4337

0.6877

Deeplabv3 + Xception [37]

0.6509

0.5385

0.6251

0.7947

Deeplabv3 + Mobile [37]

0.6303

0.4825

0.5957

0.7173

HRNetSmallv2 [40]

0.6428

0.5513

0.6811

0.7253

HRNet [40]

0.7901

0.6953

0.8796

0.7694

MSRF-Net [35]

0.7921

0.6498

0.9001

0.7000

Our method

0.8462

0.7584

0.8364

0.8681