Skip to main content
. 2025 Sep 1;15:32192. doi: 10.1038/s41598-025-18107-6

Table 2.

Normalized results of human evaluation based on perceptual similarity to the source and overall realism.

Method Similarity (%) ↑ Realism (%) ↑
CatV2TON (2025)29 39.00% 32.16%
LaDI-VTON (2023)10 67.88% 83.81%
HR-VITON (2022)26 72.63% 76.84%
Ours 77.97% 80.34%

Higher scores reflect better perceived quality.