Table 2 Ablation study on the Style Encoder and Sketch Encoder within the DCADif framework

From: DCADif: decoupled conditional adaptive time-dynamic fusion diffusion inpainting of traditional Chinese mural paintings

Style/Sketch

PSNR↑

SSIM↑

LPIPS↓

SwinStyle/CLIP

21.887

0.821

0.043

CLIP/CLIP

21.455

0.875

0.028

SwinStyle/SwinStyle

24.596

0.922

0.019

CLIP/SwinStyle

26.942

0.925

0.013

  1. In the first column of the table, the encoder configurations are specified as (Style Encoder / Sketch Encoder). The ablation study was conducted with a mask ratio of 20%-30%. The output images of the generators are used for metrics computation. ↑ Higher values are better, ↓ Lower values are better. *Optimal results are displayed in bold.
Back to article page