ART-VITON uses residual prior initialization and artifact-free measurement-guided sampling with data consistency, frequency correction, and periodic denoising to generate artifact-free virtual try-on images on VITON-HD, DressCode, and SHHQ-1.0.
˜ϵt := p 1−¯αt−1 −η 2β2 t ·ϵ θ +ηβ t ·ϵ√1−¯αt−1 ,ϵ∼ N(0,I),(16) whereηcontrols the noise level andβ t is the noise schedule
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
ART-VITON: Measurement-Guided Latent Diffusion for Artifact-Free Virtual Try-On
ART-VITON uses residual prior initialization and artifact-free measurement-guided sampling with data consistency, frequency correction, and periodic denoising to generate artifact-free virtual try-on images on VITON-HD, DressCode, and SHHQ-1.0.