HITS-AIN/IllustrisTNG_SKIRT_SDSS

IllustrisTNG SKIRT SDSS

Preprocessed synthetic galaxy images derived from the IllustrisTNG cosmological simulations. Raw multi-band FITS images were produced with the SKIRT Monte Carlo radiative transfer code in SDSS photometric bands and subsequently processed into 128 × 128 RGB images ready for machine-learning applications.

The dataset is designed as training data for the Spherinator / HiPSter framework, but it is general-purpose and suitable for any galaxy-morphology task.

---

Dataset Details

Dataset Description

Each sample represents a single galaxy subhalo rendered as an observer-frame RGB image.

Starting from the raw FITS images a deterministic preprocessing pipeline is applied:

Step	Description
RGB colour creation	Combines four SDSS bands into three channels with per-channel scalers and an arcsinh stretch
Unhealthy-data filter	Removes images containing NaN / Inf values or constant pixel content
Truncation filter	Discards galaxies whose major axis is larger than 40 % of the image extent
Horizontal alignment	Rotates each galaxy so that the morphological major axis is horizontal
Inclination filter	Discards galaxies whose apparent minor to major axis ratio is < 0.5 (near face-on)
Crop	Centre-crops to 50 % of the image extent
Resize	Bicubic resize to 128 × 128 pixels
Reflectional invariance	Canonicalises handedness so the bright side is always on the left
Gaussian blur	3 × 3 kernel smoothing
Circular mask	Pixels outside the inscribed circle are set to zero
Min–max normalisation	Pixel values rescaled to [0, 1] per image

• Curated by: HITS-AIN
• Source simulation: IllustrisTNG public data release
• Details on the production of the images: https://www.tng-project.org/data/docs/specifications/#sec5l
• Original publication (for the images): https://ui.adsabs.harvard.edu/abs/2019MNRAS.483.4140R/abstract
• Radiative transfer: SKIRT
• Preprocessing tool: PEST
• License: Creative Commons Attribution 4.0 International (CC BY 4.0)

---

Dataset Structure

Data Fields

Field	Type	Description
image	Image (128 × 128, RGB, uint8)	Preprocessed galaxy image
simulation	string	IllustrisTNG run identifier (e.g. TNG100-1)
snapshot	int32	Simulation snapshot number
subhalo_id	int32	Subhalo ID within the Friends-of-Friends group catalogue

Data Splits

The dataset ships as a single train split.

---

Dataset Creation

Source Data

Raw FITS files were downloaded from the IllustrisTNG public data release. Each file contains a multi-band image cube with four SDSS photometric bands (u, g, r, i) generated by SKIRT radiative transfer post-processing of IllustrisTNG snapshots.

File-path convention used to extract metadata:

<root>/<simulation>/sdss/snapnum_<snapshot>/data/broadband_<subhalo_id>.fits

Data Collection and Processing

Preprocessing was performed with PEST using the pipeline defined in pipelines/illustris_skirt.yaml.

---

Citation

If you use this dataset, please cite the IllustrisTNG public data release and the PEST/Spherinator paper:

@article{Nelson2019,
  author  = {Nelson, D. and Springel, V. and Pillepich, A. and others},
  title   = {The IllustrisTNG simulations: public data release},
  journal = {Computational Astrophysics and Cosmology},
  year    = {2019},
  volume  = {6},
  pages   = {2},
  doi     = {10.1186/s40668-019-0028-x}
}

@article{Polsterer2024,
  author = {Polsterer, Kai Lars and Doser, Bernd and Fehlner, Andreas and Trujillo-Gomez, Sebastian},
  title  = {{Spherinator and HiPSter: Representation Learning for Unbiased Knowledge Discovery from Simulations}},
  url    = {https://arxiv.org/abs/2406.03810},
  doi    = {10.48550/arXiv.2406.03810},
  year   = {2024}
}