HitsAndTracksPlotter

Interactive 3D visualization of CMS sim/reco using plotly

Relies on NanoHGCML ntuples. Instructions for generating these can be found here

The code is a restructured version of this repository, now using dash to make an interactive dashboard.

Setup

Setup the necessary pip environment with

pip install -r requirements.txt

Then . ./env/bin/activate before running (needed every time you open a new session)

Running is most convenient from your local computer, using lxplus etc you would need to play with tunneling or open the browser from lxplus.

For this, you could copy output files locally using cernbox. The files in the central repo can be found at

/eos/user/n/nchernya/ML4Reco/Gun50/thresholds

and

/eos/user/k/kelong/ML4Reco/Ntuples/

For example, the 50 particle sample can be copied as

cp /eos/cms/store/user/kelong/ML4Reco/Gun50Part_CHEPDef_NoPropagate/111_nanoML.root /eos/user/<yourcernbox>/ML4Reco/Ntuples

Running

Two options are available : run an interactive dashboard (mode = interactive) or run on a particular event and save an event display in .html format (mode = output)

Interactive dashboard

Launch the server with

python3 runDash.py interactive

and direct your browser to localhost:3389

Run and save event display

python3 runDash.py output 

The following options are available :

-d : input dataset .root file

-e : event number to display

-o : output file

--outDir : output directory where to save the event display

Use

This is a WIP, so you may find some that not all options work properly for all samples. Let me know if this is the case for you.

The goal is to visualize target clusters and tracks as well as the candidates we reconstruct. Several drop down menus and check boxes allow you to control what information is displayed.

Particles per endcap modifies the data set being used. In prinicple other samples could be produced, but the current options are generated with a particle Gun shot towards the HGCAL volume with 50 particles chosen randomly from muons, electrons, pions, and kaons.

Hit types selects which hits to display. Nothing prevents you from displaying both SimHits and RecHits, but it's a bad idea, because they will be on top of each other and things will run slower for no reason.

Particle trajectories show the trajectories of true particles or candidates. The trajectories are meant to guide you to the hits that should be collected for the recontruction, they are extrapolated based on the vertex, four vector and charge, and assuming a B = 3.8 T uniform field.

Hit color mode Color hits based

• their PDG ID, based on association to SimTracks and the SimTrack pdgId
• based on the candidate or SimCluster they are associated to. NOTE: Here MergedSimClusters are produced in the NanoAOD production step by merging SimClusters, produced with the fineCalo mode, based on deltaR

Particle color mode Color particles by index in the collection or by pdgId of the candidate. In the case of coloring by index, the hit index will match the particle index if the association is the same, e.g., by CaloParticles or by candidates

Show SimClusters Show the position of default SimClusters (produced in CMSSW) or the MergedSimClusters produced with the dR merging