LArbys/MPID_pytorch

PyTorch version for MPID

MPID(PyTorch): Multiple Particle Identification, PyTorch version

PyTorch version for MPID. Rewritten from the TensorFlow version for the purpose of production on Fermilab GRID Multiple Particle Identification CNN. MPID taks a 512x512 LArTPC images and returns probablities of having particles from proton, electron, muon, muon and charged pions.

Dependecies:

LArCV, ROOT, PyTorch

Setup:

0. Setup LArCV using Wiki,
1. git clone https://github.com/ruian1/MPID_pytorch.git
2. source setup.sh

Training:

0. Edit training configures in under ./uboone/run_train.sh
1. source ./uboone/run_train.sh

Inference:

0. cd ./uboone
1. python mpid_inference.py image2d.root (for image2d input).
2. python inference_pid_torch.py pixel2d.root (for pixel2d input).