Physics Datasets
Datasets for machine learning in physics, computational fluid dynamics, and scientific simulation.
CFD & PDE Benchmarks
| Dataset | Description | Link |
|---|---|---|
| PDEBench | Comprehensive benchmark for PDE solving with ML | github |
| PDEArena | PDE modeling benchmark suite | microsoft |
| BLASTNet | 744 full-domain samples of 3D turbulent flows | github |
| JHTDB | Johns Hopkins Turbulence Database | jhtdb |
| Airfoil CFD | 2D compressible flow simulations (6K samples) | zenodo |
| DrivAerNet | 4,000 car meshes with aerodynamic data | github |
Weather & Climate
| Dataset | Description | Link |
|---|---|---|
| WeatherBench 2 | ML weather forecasting benchmark | github |
| ERA5 | Global atmospheric reanalysis data | ecmwf |
| ClimSim | Climate simulation dataset for ML | github |
Particle & High-Energy Physics
| Dataset | Description | Link |
|---|---|---|
| CERN Open Data | LHC collision data and simulations | opendata.cern.ch |
| CaloGAN | Deep generative models for calorimeter simulations | ml4sci |
| LHC Olympics | Anomaly detection challenge dataset | ml4sci |
| Inference with DCTR | Direct comparison to reference for inference | ml4sci |
| Unfolding with OmniFold | ML-based unfolding for particle physics | ml4sci |
| Kaggle HEP | Particle physics ML challenges | kaggle |
Cosmology
| Dataset | Description | Link |
|---|---|---|
| CosmoFlow | ~10,000 cosmological N-body dark matter simulations | ml4sci |
Astrophysics
| Dataset | Description | Link |
|---|---|---|
| SDSS | Sloan Digital Sky Survey imaging and spectra | sdss |
| NASA Exoplanet Archive | Confirmed exoplanets and candidates | nasa |
Simulation Datasets
| Dataset | Description | Link |
|---|---|---|
| MeshGraphNets Data | DeepMind simulation datasets | deepmind |
| PhiFlow Examples | Physics simulation framework with data | github |
Dataset Collections
| Resource | Description |
|---|---|
| awesome-matchem-datasets | Materials & chemistry datasets (Blaiszik) |
| awesome-scientific-machine-learning | SciML resources |
| awesome-pinn | Physics-informed neural networks |