3D模拟数据集

本研究创建了一个名为3D模拟数据集的大型合成数据集，包含3801种不同身高和体重的人体类型，穿着45612种不同类型的衣物。数据集通过3D物理模拟器Blender生成，涵盖了多种衣物类型和尺寸，以及详细的人体属性。创建过程中，首先使用SMPL-X Blender插件生成不同高度和重量的3D人体模型，然后根据实际设计制作衣物缝制图案，并将这些图案变形应用到人体网格上，以生成穿着衣物的人体图像。此数据集旨在支持虚拟试衣系统的发展，特别是在线服装购物中尺寸和合身问题的解决，提供更真实和个性化的试衣体验。

This study constructs a large-scale synthetic dataset named 3D Simulation Dataset, which includes 3801 human body variants with varying heights and weights, wearing 45,612 distinct clothing styles. Generated via the 3D physics simulator Blender, this dataset covers diverse clothing types and sizes as well as detailed human body attributes. During the dataset creation process, the SMPL-X Blender plugin is first utilized to generate 3D human body models with different heights and weights. Subsequently, clothing sewing patterns are developed based on actual garment design specifications, and these patterns are warped and applied to the human meshes to produce images of humans wearing clothing. This dataset is intended to support the development of virtual fitting room systems, particularly addressing the size and fit issues in online clothing shopping, and deliver a more realistic and personalized fitting experience.