This dataset is used in our paper: "Generalizing to new geometries with Geometry-Aware Autoregressive Models (GAAMs) for fast calorimeter simulation". The paper is accepted by JINST, and the preprint can be found at: https://arxiv.org/abs/2305.11531.

The dataset is store in a HDF5 file: "caloflow_data_differentShapes_AllLayers.h5", which can be opened using a Pythonic interface -- h5py package. The file size is 281 MB. This dataset contains 11300 events, and each event has three layers (0, 1, 2). There are 7 features for each calorimeter image: f['Event_{group_id}_{sample_id}][{layer_id}].
1. 'EtaX': dimensionality of the eta direction
2. 'PhiY': dimensionality of the phi direction
3. 'dEta': the smallest cell size for eta direction
4. 'dPhi': the smallest cell size for phi direction
5. 'data': the calorimeter image
6. 'xEdges': the cell edge locations for the eta direction
7. 'yEdges': the cell edge locations for the phi direction

Junze Liu
Nov, 2023