OrcaSong: Preprocessing KM3NeT data for DL
The documentation for OrcaSong can be found at https://ml.pages.km3net.de/OrcaSong!
OrcaSong is a part of the Deep Learning efforts of the neutrino telescope KM3NeT. Find more information about KM3NeT on http://www.km3net.org.
In this regard, OrcaSong is a project that preprocesses raw KM3NeT detector data for the use with deep neural networks, making use of km3nets data processing pipline km3pipe. Two different modes are available:
- For convolutional networks: produce n-dimensional 'images' (histograms)
- For graph networks: produce a list of nodes, each node representing infos about a hit in the detector
Currently, only simulations with a hdf5 data format are supported as an input.
OrcaSong can be installed via pip by running:
pip install orcasongA Singularity image of the latest stable version of OrcaSong is also provided.
You can download it from the km3net sftp server pi1139.physik.uni-erlangen.de
in singularity/orcasong.sif.