DPNNet (Synthetic Protoplanetary Disk Images)

This Zenodo repository contains the protoplanetary disk image datasets used in the studies on machine-learning-based inference and image modeling presented in the VADER and DPNNet papers. The purpose of this repository is to provide the exact datasets used in the experiments to ensure full reproducibility of the published results and to facilitate reuse by the community working on disk–planet interaction and machine learning in astrophysics.

📦 Contents of the Repository

The repository includes:

1. Synthetic Protoplanetary Disk Images
High-resolution synthetic images of protoplanetary disks generated using hydrodynamic simulations FARGO3D, followed by radiative-transfer (using RADMC3D) post-processing. The images cover a wide range of disk morphologies, including:

• Gaps and rings due to multi-planetary systems
• Planet-induced structures at various viewing angles

These images are the ones used to train and evaluate the models in both the VADER and DPNNet studies.

Labels and Physical Parameters

Each image is accompanied by metadata describing the physical setup used to generate it. This includes (where applicable):

• Planet mass
• Planet location
• Disk physical parameters
• Viewing geometry (inclination, position angle, etc.)
• Radiative-transfer settings

The labels are stored in structured files so that the dataset can be directly used for supervised machine-learning applications.

For metadata, please contact the authors of the paper.

📚 Associated Publications

This dataset is associated with the following works:

• The VADER paper (https://iopscience.iop.org/article/10.3847/1538-4357/ae22ed)
• The DBNNet-3.0 paper

Please cite both the Zenodo dataset and the associated publications when using this dataset in scientific work.