DrivAerNet++
收藏资源简介:
DrivAerNet++是由麻省理工学院机械工程系创建的一个大规模多模态汽车数据集,包含8000种高保真计算流体动力学(CFD)模拟的汽车设计。数据集涵盖了快背、斜背和旅行车背等多种车型,以及不同的底盘和车轮设计,代表内燃机和电动汽车。每个数据集条目都包含详细的3D网格、参数模型、空气动力学系数以及广泛的流动和表面场数据,以及用于汽车分类和点云数据的分割部件。该数据集支持数据驱动的设计优化、生成建模、代理模型训练、CFD模拟加速和几何分类等多种机器学习应用。DrivAerNet++通过提供超过39TB的公开工程数据,填补了可用资源的显著缺口,为模型训练、推广和加速汽车设计过程提供了高质量、多样化的数据。
DrivAerNet++ is a large-scale multimodal automotive dataset developed by the Department of Mechanical Engineering at the Massachusetts Institute of Technology, encompassing 8,000 high-fidelity Computational Fluid Dynamics (CFD) simulated vehicle designs. The dataset covers various vehicle body configurations including fastback, liftback, and station wagon styles, alongside diverse chassis and wheel designs, representing both internal combustion engine (ICE) and electric vehicles (EVs). Each dataset entry contains detailed 3D meshes, parametric models, aerodynamic coefficients, extensive flow and surface field data, as well as segmented components for vehicle classification and point cloud data applications. This dataset supports a wide range of machine learning applications such as data-driven design optimization, generative modeling, surrogate model training, CFD simulation acceleration, and geometric classification. By providing over 39 terabytes of publicly available engineering data, DrivAerNet++ fills a significant gap in existing resources, delivering high-quality, diverse data for model training, generalization, and acceleration of the automotive design process.

- 1DrivAerNet++: A Large-Scale Multimodal Car Dataset with Computational Fluid Dynamics Simulations and Deep Learning Benchmarks麻省理工学院机械工程系 · 2024年



