Fast calorimeter simulation

Tools and studies for fast simulation of calorimeter showers: detailed Geant4 / DD4hep simulation, compact point-cloud representations of the showers, and their use as input to machine-learning-based fast simulation. This organization hosts the trained weights; the code is on GitHub.

One 10 GeV photon shower in the Open Data Detector: 3,604 Geant4 steps, 365 points after merging within cells, 266 points after HDBSCAN clustering
One photon shower as Geant4 steps and after two step2point compressions.

Publications, code and data

Paper Code Data and weights
step2point dataset: Detailed shower simulation for data representation studies, arXiv:2509.22340 (CDS) step2point Zenodo, doi:10.5281/zenodo.17199427

Library documentation: fast-sim.github.io/step2point