Virtual Try-On
Overview
Virtual try-on studies how to place garments onto people in images and videos, with uses in online shopping, outfit design, and personalized avatars. Most systems use diffusion or flow-matching models conditioned on garment references, often adding masks, pose or layout cues, text instructions, or learned visual correspondences; 3D avatars and video models support changes in viewpoint and motion. Evaluation commonly tests garment fidelity, body and background preservation, fit, and temporal consistency on paired and in-the-wild benchmarks, increasingly alongside human or multimodal judgments. A central challenge is balancing accurate garment shape and texture with natural draping, pose and identity preservation, flexible control, and efficient generation across diverse items and settings.
Latest papers
55 papers2 papers in the last four weeks, down 60% on the four weeks before. 0.0% of all new papers.