Skip to content
SparseEventID for SBND Events
Jupyter Notebook Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
larcv_scripts Updated preprocessing files. Aug 1, 2019

Sparse Event ID

This repository contains a set of models that work on sparse data in pytorch. There are several dependencies for some models (Like Submanifold sparse convolutions)

In general, all models share the same tools for IO, distributed training, saving and restoring, and doing evaluations of test, train, and inference steps.

The models included here are (initially):


Implementation of PointNet for event identification in neutrino events in lartpcs. Implementation is available for both 2D and 3D networks.

Submanifold sparse convolution ResNet

Implementation of a standard resnet architecture with submanifold sparse convolutions. Available for 2D and 3D. In 2D, its possible to use the multiplane archicture with shared weights across planes, which merges downstream.

(Coming eventually)

Dynamic Graph Convolutional Neural Network


  • The IO model is based on larcv, so you need larcv data files to run this.
  • The sparse convolution requires the submanifold sparse convolution functions, from facebook.
  • Pytorch is the neural network framework
  • To run in a distrubuted way, horovod and mpi4py are required.



bin/ train [options]

to execute the point net algorithm. Similarly, you can run pointnet_seg, dgcnn, dgcnn_seg.

You can’t perform that action at this time.