COCOA

AI algorithms need training data — high-quantity, high-quality data. As such, developing deep learning models for collider physics requires data from accurate detector simulators. We created COCOA to address the lack of available tools in this area, besides a handful of highly complex detector simulations. COCOA is a user-configurable multi-layer calorimeter simulation based on GEANT4 with an additional simple tracking emulator. Users can easily simulate particle trajectories in a realistic, LHC-like detector environment and obtain the truth record of which particles deposited energy where. Datasets simulated with COCOA have enabled several of our projects applying deep learning to low-level reconstruction and generation tasks.

Links:

• Paper: https://iopscience.iop.org/article/10.1088/2632-2153/acf186
• GitHub: https://github.com/cocoa-hep/cocoa-hep