Skip to main content

Table 8 Ablation study of EG-Transunet on the Kvasir-SEG

From: EG-TransUNet: a transformer-based U-Net with enhanced and guided models for biomedical image segmentation

Method

mDice

mIoU

Recall

Precision

Flops

EG-N w/o PEM + CSA + SGA

0.8820

0.8269

0.8956

0.8990

16.3G

EG-N w/o PEM + SGA

0.9248

0.8796

0.9342

0.9368

16.9G

PAS w/o PEM + CSA

0.9227

0.8790

0.9258

0.9336

21.8G

EG-N w/o SGA + CSA

0.9263

0.8815

0.9303

0.9345

18.8G

EG-N w/o PEM

0.9284

0.8754

0.9388

0.9281

22.2G

EG-N w/o SGA

0.9301

0.8867

0.9363

0.9404

19.3G

EG-N w/o CSA

0.9334

0.8849

0.9369

0.9353

24.4G

EG-TransUNet

0.9344

0.8927

0.9401

0.9436

24.9G

  1. For each column, the best results are highlighted