This study explores the use of physics-based cloth simulation tools (e.g., Unity, Blender) to generate high-quality synthetic datasets for fashion applications. By simulating garments under various conditions—different poses, occlusions, and lighting—one can create diverse and challenging training data. The resulting dataset can be leveraged to enhance model performance on downstream tasks such as Virtual Try-On (VTON) and Virtual Try-Off (VTOFF) while also serving as a benchmark for evaluating model robustness in real-world scenarios.
Literature
Simulators:
Similar datasets: