ShapeNet Car Computational Fluid Dynamics Simulations

20 simulation samples for Computational Fluid Dynamics analysis, split into 10 training and 10 test samples. The dataset was generated using CFD physics simulation and converted to PLAID format for standardized access by the author fabiencasenave.

Use Cases

• Train a neural operator model on CFD simulation data to predict aerodynamic flow fields around car geometries.
• Analyze aerodynamic performance across different ShapeNet car designs using simulation outputs.
• Benchmark physics-informed machine learning models on a standardized 20-sample CFD dataset.

Strengths

• Contains 20 total simulation samples with a defined 10/10 train/test split.
• Data is generated from physics-based CFD simulations, providing a synthetic ground truth.
• Stored in the standardized PLAID format (version 0.1.13.dev31+gda966e466) for consistent access.

Limitations

• Very small sample size of only 20 total simulations, limiting statistical power.
• Column names and data structure are unknown, requiring inspection of the PLAID format documentation.
• Sample data and file size are unavailable for preview, complicating initial assessment.

Provenance

Source

NeuralOperator via Zenodo (https://zenodo.org/records/13993629)

Collection Method

Physics-based CFD simulation data converted to PLAID format.

Freshness

Last updated on March 21, 2026.