v2.8.1

Includes CRT-TPC matching and important bug fixes.

v2.8

Last major release before v3.0:

• Major reconstruction chain execution speed improvements
• HDF5 writer fully integrated
• Complete overhaul of the analysis tools

2.7.1

Bug fixes in the analysis tools, various speedup.

v2.7

This version includes breaking changes (PID weights trained prior to this will be broken) and should fix a number of bugs.
Also code cleanup of arxiv folders and deprecated parsers removed. See PR#123 for details.

v2.6.1

See PR#122 for details

v2.6

Description to be updated

v2.5.1

To enable running on neutrino MC samples:

• Include fixes for empty fragments/particles list in Grappa (Francois) and analysis tools (Temigo)

v2.5

• Recent changes to adapt_labels are included (assign label -1 to ambiguous cases to avoid distance-based or other heuristics in deciding truth labels for true ghost voxels predicted as non-ghost, also adjust cluster id vs group id for track-like particles that are broken after deghosting).
• GraphSpice and Grappa clustering targets have been improved (fixes #111 for example)
• Enforce PID predictions + semantic agreement

Configurations that are good picks for this release are

• inference_full_volume_072022.cfg (older dataset, trained with improved labels)
• inference_full_volume_082022.cfg (newer dataset with improved particle kinematics ranges + trained with improved labels)

v2.4.2

• Fixes to be able to run on neutrino samples (mostly edge cases). Note for future: make sure sample is filtered (removing entries with empty pcluster information).
• Updated vertex heuristic parameters to most recent ones from last study

v2.4.1

• Updated documentation HowTo for larcv::Neutrino access
• depositions units clean-up in analysis tools
• Fix bug in stopping muon selection (was taking wrong endpoint occasionally because of indexing mistake)