DrivAerNet++

DrivAerNet++ extends the classic DrivAer benchmark to 8,000 diverse parametric car geometries, each paired with high-fidelity CFD results: 3D flow fields, surface fields, meshes and drag/lift coefficients.

Use cases: Training geometry-to-field surrogate models; aerodynamic coefficient regression; generative design and shape optimization.

The accompanying paper documents dataset construction, mesh generation and baseline benchmarks for common surrogate architectures.