Return
Virtual try-on algorithm integrating multi-view user perspectives with multi-level sampling warping module (M-LSWM)
X
H
Y
X
DOI:10.1016/j.asoc.2026.116179.png)
Abstract
En 中文
• A multi-view input strategy is proposed to capture spatial information from multiple angles, solving the problem of insufficient single-view spatial perception. • The M-LSWM module precisely handles local garment deformation, preserving fine textures and edges while ensuring semantic consistency. • A BoTNet-based generator with multi-head self-attention reduces computational redundancy and enhances local detail extraction.
Journal
IF:
6.6
Papers:
1.4W
Citations:
4.8W
