Super-resolution for calorimeters

Generative AI pipelines often upsample image resolution by upsampling the number of pixels based on learned features. We are exploring analogous super-resolution techniques for particle detectors, where calorimeter cells play the role of pixels. In [1], we were the first to demonstrate the potential of this approach using a simplified detector simulation. In [2], we extended the study to a more realistic dataset and incorporated state-of-the-art deep learning methods, showing that super-resolution can significantly enhance reconstruction quality and reduce noise.

[1] https://link.springer.com/article/10.1140/epjc/s10052-021-08897-0 
[2] https://arxiv.org/abs/2409.16052