Table 7 Ablation experiments of different components.

From: A structural information-guided cross-modal method for damaged inscription inpainting via vision-language models

Methods

PSNR ↑ 

SSIM ↑ 

LPIPS ↓ 

FID ↓ 

StSc ↑ 

CINet

21.7686

0.9564

0.0370

16.2541

0.9534

CINet-FSM

21.6633

0.9551

0.0388

19.7553

0.9476

CINet-Semb

21.3994

0.9534

0.0382

16.6070

0.9301

CINet-ETEX

21.5526

0.9551

0.0381

18.8814

0.9505

CINet-CAIMG

21.7001

0.9550

0.0373

16.2934

0.9491

CINet-CATEX

21.4337

0.9438

0.0389

16.4120

0.9389

CINet-CATEX-CAIMG

21.3917

0.9533

0.0382

16.5173

0.9345

CINet-FSM- CATEX-CAIMG

21.6970

0.9483

0.0387

26.5635

0.9461

CINet-FSM- CATEX-CAIMG-Semb

21.5443

0.9216

0.0386

32.5536

0.9389

CINet-FSM- CATEX-CAIMG-Semb-ETEX

20.5643

0.9335

0.0431

37.5393

0.9127

  1. The optimal results are shown in bold, and the sub-optimal results are underlined.