Improve WindsorML dataset card

Add structured metadata, an at-a-glance summary, clearer navigation and citation guidance, and preview-first Hugging Face CLI download examples. Preserve the benchmark splits, current asset-coverage caveat, native surface-area documentation, acknowledgements, license, and version history.

README.md

@@ -1,61 +1,82 @@
 ---
+pretty_name: WindsorML
 license: cc-by-sa-4.0
+viewer: false
+tags:
+- computational-fluid-dynamics
+- scientific-machine-learning
+- automotive-aerodynamics
+- high-fidelity-simulation
+- surrogate-modeling
+- ai-for-science
+- open-data
 ---
-WindsorML: High-Fidelity Computational Fluid Dynamics dataset for automotive aerodynamics
--------
 
-Contact: 
-----------
-Neil Ashton (NVIDIA) - contact@caemldatasets.org
+# WindsorML: High-Fidelity Computational Fluid Dynamics Dataset for Automotive Aerodynamics
 
-website:
-----------
-https://caemldatasets.org
+[Project website](https://caemldatasets.org) · [Paper](https://proceedings.neurips.cc/paper_files/paper/2024/hash/42a59a5f35b1b3c3fd648397c88a7164-Abstract-Datasets_and_Benchmarks_Track.html) · [Hugging Face paper page](https://huggingface.co/papers/2407.19320) · [Official splits](splits/)
 
-Summary:
--------
+## At a Glance
 
-This work presents a new open-source high-fidelity dataset for Machine Learning (ML) containing 355 geometric variants of the Windsor body, to help the development and testing of ML surrogate models for external automotive aerodynamics. Each Computational Fluid Dynamics (CFD) simulation was run with a GPU-native high-fidelity Wall-Modeled Large-Eddy Simulations (WMLES) using a Cartesian immersed-boundary method using more than 280M cells to ensure the greatest possible accuracy. The dataset contains geometry variants that exhibits a wide range of flow characteristics that are representative of those observed on road-cars. The dataset itself contains the 3D time-averaged volume & boundary data as well as the geometry and force & moment coefficients. This paper discusses the validation of the underlying CFD methods as well as contents and structure of the dataset. To the authors knowledge, this represents the first, large-scale high-fidelity CFD dataset for the Windsor body with a permissive open-source license (CC-BY-SA).
+| | |
+|---|---|
+| Domain | External automotive aerodynamics |
+| Scale | 355 geometry variants; per-run assets are currently available for 350 cases |
+| CFD method | GPU-native wall-modelled LES with a Cartesian immersed-boundary method |
+| Simulation size | More than 280 million cells and approximately 80 convective time units per case |
+| Published data | Surface and volume fields, geometry, images, and force/moment coefficients |
+| License | CC BY-SA 4.0 |
 
-CFD Solver:
-----------
-All cases were run using the Volcano Platforms commerical CFD solver, which is based upon a GPU-native cartesian immersed-boundary method Wall-Modelled Large-Eddy Simulation (WMLES) approach. Each case was run transiently for approximately 80 convective time units (CTU) on meshes of approximately 300M cells. Please see the paper for full details on the code and validation:
+## Summary
 
-How to cite this dataset:
-----------------
-In order to cite the use of this dataset please cite the paper below which contains full details on the dataset. It can be found here: https://arxiv.org/abs/2407.19320
+WindsorML is an open high-fidelity CFD dataset containing 355 geometric variants of the Windsor body for developing and testing machine-learning surrogate models in external automotive aerodynamics. Each simulation used a GPU-native wall-modelled Large-Eddy Simulation (WMLES) approach with a Cartesian immersed-boundary method and more than 280 million cells.
 
-@article{ashton2024windsor,
-    title = {WindsorML: High-Fidelity Computational Fluid Dynamics dataset for automotive aerodynamics},
-    year = {2024},
-    journal = {arxiv.org},
-    url={https://arxiv.org/abs/2407.19320},
-    author = {Ashton, Neil and Angel, Jordan and Ghate, Aditya and Kenway, Gaetan and Long Wong, Man and Kiris, Cetin and Walle, Astrid and Maddix, Danielle and Page, Gary}
+The geometry variants exhibit a wide range of flow characteristics representative of road vehicles. The dataset provides three-dimensional time-averaged volume and boundary data, geometry, images, and force and moment coefficients. The accompanying paper documents the CFD validation, dataset contents, and structure.
+
+## CFD Solver
+
+All cases were run with the Volcano Platforms commercial CFD solver, based on a GPU-native Cartesian immersed-boundary WMLES approach. Each transient simulation covered approximately 80 convective time units (CTU) on a mesh of approximately 300 million cells. See the [paper](https://proceedings.neurips.cc/paper_files/paper/2024/hash/42a59a5f35b1b3c3fd648397c88a7164-Abstract-Datasets_and_Benchmarks_Track.html) for full solver and validation details.
+
+## How to Cite This Dataset
+
+Please cite the corresponding NeurIPS Datasets and Benchmarks paper:
+
+```bibtex
+@inproceedings{ashton2024windsor,
+  title     = {{WindsorML: High-Fidelity Computational Fluid Dynamics Dataset for Automotive Aerodynamics}},
+  author    = {Ashton, Neil and Angel, Jordan and Ghate, Aditya and Kenway, Gaetan and Wong, Man Long and Kiris, Cetin and Walle, Astrid and Maddix, Danielle and Page, Gary},
+  booktitle = {Advances in Neural Information Processing Systems},
+  volume    = {37},
+  pages     = {37823--37835},
+  year      = {2024},
+  url       = {https://proceedings.neurips.cc/paper_files/paper/2024/hash/42a59a5f35b1b3c3fd648397c88a7164-Abstract-Datasets_and_Benchmarks_Track.html}
 }
+```
+
+## Dataset Structure and Files
+
+Each `run_i` folder corresponds to a different geometry and contains the following files, where `i` is the run number:
+
+* `windsor_i.stl`: Geometry STL (approximately 5 MB).
+* `windsor_i.stp`: Geometry STEP file (approximately 1 MB).
+* `geo_parameters_i.csv`: Parameters defining the geometry, as documented in the paper.
+* `boundary_i.vtu`: Surface boundary data (approximately 500 MB).
+* `boundary_dual_area_i.npy`: Native-point surface quadrature weights in `m^2`, in the exact point order of `boundary_i.vtu`.
+* `volume_i.vtu`: Volume field data (approximately 20 GB).
+* `force_mom_i.csv`: Time-averaged force and moment coefficients (`Cd`, `Cs`, `Cl`, `Cmy`).
+* `force_mom_varref_i.csv`: Time-averaged coefficients using a unique reference area for each geometry.
+* `images/`: Geometry and flow-variable images, including pressure, streamwise velocity, and normal Reynolds stresses on x-, y-, and z-oriented slices.
 
-Files:
--------
-Each folder (e.g run_1,run_2…run_“i” etc) corresponds to a different geometry that contains the following files where “i” is the run number:
-
-* windsor_i.stl : geometry stl (~5mb)
-* windsor_i.stp : geometry step (~1mb)
-* geo_parameters_1.csv : parameters that define the geometry (explained in the associated paper)
-* boundary_i.vtu : Boundary VTU (~500mb)
-* `boundary_dual_area_i.npy`: Native-point surface quadrature weights (`m^2`) for `boundary_i.vtu`, in exact point order.
-* volume_i.vtu : Volume field VTU (~20GB)
-* force_mom_i.csv : forces/moments time-averaged (Cd,Cs,Cl,Cmy)
-* force_mom_varref_i.csv: forces/moments time-averaged (Cd,Cs,Cl,Cmy) using unique reference area per geometry
-* images (folder) that contains images of the following variables (pressure, velocityX,ReynoldsStressXX,YY,ZZ) for slices of the domain in the X,Y & Z locations as well as an image of the geometry itself (windsor_i.png)
-* force_mom_all.csv: contains force/moments for all runs in a single file
-* force_mom_varref_all.csv: contains force/moments for all runs in a single file using a reference frontal area that is unique to each geometry
-* geo_parameters_all.csv: contains all the geometry parameters for all the runs in a single file
-* [`splits/`](splits/): deterministic benchmark manifests, methods documentation, derived metrics, diagnostic figures, and generation code
-
-## Recommended dataset splits
-
-WindsorML provides eight deterministic train/validation/test split families in
-[`splits/manifest.json`](splits/manifest.json). Identifiers follow the
-`run_N` convention used by the dataset.
+The repository root also contains:
+
+* `force_mom_all.csv`: Force and moment coefficients for all runs in one file.
+* `force_mom_varref_all.csv`: Coefficients for all runs using a geometry-dependent reference frontal area.
+* `geo_parameters_all.csv`: Geometry parameters for all runs in one file.
+* [`splits/`](splits/): Deterministic benchmark manifests, methods documentation, derived metrics, diagnostic figures, and generation code.
+
+## Recommended Dataset Splits
+
+WindsorML provides eight deterministic train/validation/test split families in [`splits/manifest.json`](splits/manifest.json). Identifiers follow the `run_N` convention used by the dataset.
 
 | Split | Type | Train | Validation | Test | Intended evaluation |
 |---|---:|---:|---:|---:|---|
@@ -68,96 +89,88 @@ WindsorML provides eight deterministic train/validation/test split families in
 | `low_drag` | OOD | 248 | 36 | 71 | Low-drag extrapolation |
 | `image_wake` | OOD | 248 | 36 | 71 | Image-derived wake extrapolation |
 
-The `full` family is a reproducible seed-42 benchmark. It is not a
-reconstruction of the paper's preliminary 60/20/20 evaluation partition,
-whose case membership was not published. The reduced-data training sets are
-strictly nested and share the same validation and test cases. For the OOD
-families, validation is selected from the training-side population.
+The `full` family is a reproducible seed-42 benchmark. It is not a reconstruction of the paper's preliminary 60/20/20 evaluation partition, whose case membership was not published. The reduced-data training sets are strictly nested and share the same validation and test cases. For the OOD families, validation is selected from the training-side population.
 
-The aggregate tables and manifest cover `run_0` through `run_354`. At the
-source revision used for this package, per-run STL and image files for
-`run_350` through `run_354` were unavailable. Their geometry and image-wake
-scores are estimated from nearby observed cases and are explicitly identified
-in the distributed metric CSVs.
+The aggregate tables and manifest cover `run_0` through `run_354`. At the source revision used for this package, per-run STL and image files for `run_350` through `run_354` were unavailable. Their geometry and image-wake scores are estimated from nearby observed cases and are explicitly identified in the distributed metric CSVs.
 
-Download only the split package with:
+Complete definitions, construction methods, missing-data treatment, diagnostic figures, and reproducibility instructions are provided in [`splits/README.md`](splits/README.md).
 
-```bash
-hf download neashton/windsorml \
-  --type dataset \
-  --include "splits/**" \
-  --local-dir ./windsorml
-```
+## Native-Point Surface Areas
 
-Complete definitions, construction methods, missing-data treatment,
-diagnostic figures, and reproducibility instructions are provided in
-[`splits/README.md`](splits/README.md).
+Each `run_N/boundary_dual_area_N.npy` is a one-dimensional little-endian float32 array containing one barycentric dual-area weight in `m^2` per native point of `run_N/boundary_N.vtu`, in its original point order. The array is valid only for that exact raw VTU. The definition, generator, hashes, and complete 350-run manifest are in [`surface_dual_areas/`](surface_dual_areas/).
 
-Downloads: 
------------
+## How to Download
 
-The dataset is now available on HuggingFace. Below are some examples of how to download all or selected parts of the dataset. Please refer to the HuggingFace documentation for other ways to accessing the dataset and building workflows.
+WindsorML is publicly available on Hugging Face and is approximately 8 TB in full. We recommend the Hugging Face CLI for selective, resumable downloads and a dry-run preview before transferring large files.
 
-Example 1: Download all files (~8TB)
--------
-Please note you’ll need to have git lfs installed first, then you can run the following command:
+### Prerequisites
 
-```
-git clone git@hf.co:datasets/neashton/windsorml
+```bash
+pip install -U huggingface_hub hf_xet
 ```
 
-Example 2: only download select files (STL,images & force and moments):
--------
-Create the following bash script that could be adapted to loop through only select runs or to change to download different files e.g boundary/volume.
+Authentication is optional for this public dataset, but recommended for large downloads to avoid unauthenticated rate limits:
 
 ```bash
-#!/bin/bash
+hf auth login
+```
 
-# Set the path and prefix
-HF_OWNER="neashton"
-HF_PREFIX="windsorml"
+### Preview or Download the Full Dataset
 
-# Set the local directory to download the files
-LOCAL_DIR="./windsor_data"
+```bash
+hf download neashton/windsorml \
+  --repo-type dataset \
+  --local-dir ./windsorml_data \
+  --dry-run
+```
 
-# Create the local directory if it doesn't exist
-mkdir -p "$LOCAL_DIR"
+Remove `--dry-run` to start the download.
 
-# The currently available per-run folders span 0 to 349.
-for i in $(seq 0 349); do
-    RUN_DIR="run_$i"
-    RUN_LOCAL_DIR="$LOCAL_DIR/$RUN_DIR"
+### Preview Selected Metadata and Benchmark Files
 
-    # Create the run directory if it doesn't exist
-    mkdir -p "$RUN_LOCAL_DIR"
+```bash
+hf download neashton/windsorml \
+  --repo-type dataset \
+  --local-dir ./windsorml_data \
+  --include "force_mom_all.csv" \
+  --include "force_mom_varref_all.csv" \
+  --include "geo_parameters_all.csv" \
+  --include "splits/**" \
+  --dry-run
+```
 
-    # Download the windsor_i.stl file
-    wget "https://huggingface.co/datasets/${HF_OWNER}/${HF_PREFIX}/resolve/main/$RUN_DIR/windsor_$i.stl" -O "$RUN_LOCAL_DIR/windsor_$i.stl"
+Remove `--dry-run` to download those files.
 
-    # Download the force_mom_i.csv file
-    wget "https://huggingface.co/datasets/${HF_OWNER}/${HF_PREFIX}/resolve/main/$RUN_DIR/force_mom_$i.csv" -O "$RUN_LOCAL_DIR/force_mom_$i.csv"
+### Preview a Small Per-Run Subset
 
-done
+```bash
+hf download neashton/windsorml \
+  --repo-type dataset \
+  --local-dir ./windsorml_data \
+  --include "run_0/windsor_0.stl" \
+  --include "run_0/force_mom_0.csv" \
+  --dry-run
 ```
 
-Acknowledgements
------------
-* CFD solver and workflow development by Jordan Angel, Aditya Ghate, Gaetan Kenway, Man Long Wong, Cetin Kiris (Volcano Platforms) and Neil Ashton (Amazon Web Services - now NVIDIA)
-* Geometry parameterization by Astrid Walle (Siemens Energy)
-* Windsor advise and consultation by Gary Page (Loughborough University)
-* Guidance on dataset preparation for ML by Danielle Maddix (Amazon Web Services - now NVIDIA)
-* Simulation runs, HPC setup and dataset preparation by Neil Ashton (Amazon Web Services - now NVIDIA )
+Remove `--dry-run` to download the selected files.
 
-License
-----
-This dataset is provided under the CC BY SA 4.0 license, please see LICENSE.txt for full license text.
+## Acknowledgements
 
-## Native-point surface areas
+* CFD solver and workflow development by Jordan Angel, Aditya Ghate, Gaetan Kenway, Man Long Wong, and Cetin Kiris (Volcano Platforms), and Neil Ashton (Amazon Web Services, now NVIDIA).
+* Geometry parameterization by Astrid Walle (Siemens Energy).
+* Windsor advice and consultation by Gary Page (Loughborough University).
+* Guidance on dataset preparation for ML by Danielle Maddix (Amazon Web Services, now NVIDIA).
+* Simulation runs, HPC setup, and dataset preparation by Neil Ashton (Amazon Web Services, now NVIDIA).
 
-Each `run_N/boundary_dual_area_N.npy` is a one-dimensional little-endian float32 array containing one barycentric dual-area weight in `m^2` per native point of `run_N/boundary_N.vtu`, in its original point order. The array is valid only for that exact raw VTU. The definition, generator, hashes, and complete 350-run manifest are in [`surface_dual_areas/`](surface_dual_areas/).
+## License
+
+This dataset is provided under the CC BY-SA 4.0 license. See `LICENSE.txt` for the full license text.
+
+## Contact
+
+Dataset questions and corrections: Neil Ashton at [contact@caemldatasets.org](mailto:contact@caemldatasets.org).
 
-version history:
----------------
+## Version History
 
 * 17/08/2026 - Added native-point surface-area arrays for all 350 available runs, with reproducibility metadata.
 * 17/08/2026 - Added deterministic benchmark train/validation/test splits, including nested data-efficiency and out-of-distribution evaluation protocols; documented current per-run asset coverage.