Freedrape: Free-Form Garment Transfer Via Pose-Aware Bidirectional Warping
Journal
2025 IEEE International Conference on Image Processing Workshops, ICIPW 2025 - Proceedings
Start Page
517
End Page
522
ISBN
[9798331577995]
Date Issued
2025-09-14
Author(s)
Hsu, Chiao-Yin
Abstract
Virtual Try-On (VTON) technology has transformed online fashion by enabling users to preview outfits digitally. Traditional VTON systems rely on frontal, in-shop images, limiting their adaptability across poses. This work explores a VTON variant called apparel transfer, which allows garments to be transferred directly from one person to another. The proposed FreeDrape synthesizes realistic try-on results without requiring ground-truth supervision. FreeDrape leverages pose-aware bidirectional warping and appearance flow estimation to model how clothing deforms across varied human poses. Additional modules for shape adjustment and pattern restoration further improve realism. These components help preserve fine clothing details and structural consistency. Experimental results demonstrate that FreeDrape outperforms existing approaches, particularly in retaining intricate textures and adapting to diverse body poses, making it a robust solution for free-form garment transfer in real-world applications.
Event(s)
2025 IEEE International Conference on Image Processing Workshops, ICIPW 2025
Subjects
FreeDrape
Garment Transfer
Pose-Aware Synthesis
Virtual Try-On (VTON)
Publisher
IEEE
Type
conference paper