TICLv5a : Improve skeleton linking and parameters tuning

HLTrigger/Configuration/python/HLT_75e33/modules/hltFilteredLayerClustersRecoveryL1Seeded_cfi.py

@@ -0,0 +1,14 @@
+import FWCore.ParameterSet.Config as cms
+
+hltFilteredLayerClustersRecoveryL1Seeded = cms.EDProducer("FilteredLayerClustersProducer",
+    LayerClusters = cms.InputTag("hltHgcalMergeLayerClustersL1Seeded"),
+    LayerClustersInputMask = cms.InputTag("hltTiclTrackstersCLUE3DHighL1Seeded"),
+    algo_number = cms.vint32(6, 7, 8),
+    clusterFilter = cms.string('ClusterFilterBySize'),
+    iteration_label = cms.string('Recovery'),
+    max_cluster_size = cms.int32(9999),
+    max_layerId = cms.int32(9999),
+    mightGet = cms.optional.untracked.vstring,
+    min_cluster_size = cms.int32(2),
+    min_layerId = cms.int32(0)
+)

HLTrigger/Configuration/python/HLT_75e33/modules/hltTiclTracksterLinksL1Seeded_cfi.py

@@ -1,25 +1,12 @@
 import FWCore.ParameterSet.Config as cms
+from ..psets.hltTiclTracksterLinksPSet_cfi import hltTiclTracksterLinksPSet 
 
 hltTiclTracksterLinksL1Seeded = cms.EDProducer("TracksterLinksProducer",
     detector = cms.string('HGCAL'),
     layer_clusters = cms.InputTag("hltHgcalMergeLayerClustersL1Seeded"),
     layer_clustersTime = cms.InputTag("hltHgcalMergeLayerClustersL1Seeded","timeLayerCluster"),
     inferenceAlgo = cms.string('TracksterInferenceByDNN'),
-    linkingPSet = cms.PSet(
-        algo_verbosity = cms.int32(0),
-        cylinder_radius_sqr = cms.vdouble(9, 9),
-        dot_prod_th = cms.double(0.97),
-        max_distance_projective_sqr = cms.vdouble(60, 60),
-        max_distance_projective_sqr_closest_points = cms.vdouble(60, 60),
-        max_z_distance_closest_points = cms.vdouble(35, 35),
-        min_distance_z = cms.vdouble(30, 30),
-        min_num_lcs = cms.uint32(7),
-        min_trackster_energy = cms.double(10),
-        pca_quality_th = cms.double(0.85),
-        track_time_quality_threshold = cms.double(0.5),
-        type = cms.string('Skeletons'),
-        wind = cms.double(0.036)
-    ),
+    linkingPSet = hltTiclTracksterLinksPSet,
     pluginInferenceAlgoTracksterInferenceByDNN = cms.PSet(
         algo_verbosity = cms.int32(0),
         onnxPIDModelPath = cms.FileInPath('RecoHGCal/TICL/data/ticlv5/onnx_models/linking/id_v0.onnx'),
@@ -38,7 +25,7 @@
     original_masks = cms.VInputTag("hltHgcalMergeLayerClustersL1Seeded:InitialLayerClustersMask"),
     propagator = cms.string('PropagatorWithMaterial'),
     regressionAndPid = cms.bool(True),
-    tracksters_collections = cms.VInputTag("hltTiclTrackstersCLUE3DHighL1Seeded", "hltTiclTrackstersPassthrough")
+    tracksters_collections = cms.VInputTag("hltTiclTrackstersCLUE3DHighL1Seeded", "hltTiclTrackstersRecoveryL1Seeded")
 )
 
 

HLTrigger/Configuration/python/HLT_75e33/modules/hltTiclTracksterLinks_cfi.py

@@ -1,25 +1,12 @@
 import FWCore.ParameterSet.Config as cms
+from ..psets.hltTiclTracksterLinksPSet_cfi import hltTiclTracksterLinksPSet 
 
 hltTiclTracksterLinks = cms.EDProducer("TracksterLinksProducer",
     detector = cms.string('HGCAL'),
     layer_clusters = cms.InputTag("hltHgcalMergeLayerClusters"),
     layer_clustersTime = cms.InputTag("hltHgcalMergeLayerClusters","timeLayerCluster"),
     inferenceAlgo = cms.string('TracksterInferenceByDNN'),
-    linkingPSet = cms.PSet(
-        algo_verbosity = cms.int32(0),
-        cylinder_radius_sqr = cms.vdouble(9, 9),
-        dot_prod_th = cms.double(0.97),
-        max_distance_projective_sqr = cms.vdouble(60, 60),
-        max_distance_projective_sqr_closest_points = cms.vdouble(60, 60),
-        max_z_distance_closest_points = cms.vdouble(35, 35),
-        min_distance_z = cms.vdouble(30, 30),
-        min_num_lcs = cms.uint32(7),
-        min_trackster_energy = cms.double(10),
-        pca_quality_th = cms.double(0.85),
-        track_time_quality_threshold = cms.double(0.5),
-        type = cms.string('Skeletons'),
-        wind = cms.double(0.036)
-    ),
+    linkingPSet = hltTiclTracksterLinksPSet,
     pluginInferenceAlgoTracksterInferenceByDNN = cms.PSet(
         algo_verbosity = cms.int32(0),
         onnxPIDModelPath = cms.FileInPath('RecoHGCal/TICL/data/ticlv5/onnx_models/linking/id_v0.onnx'),

HLTrigger/Configuration/python/HLT_75e33/modules/hltTiclTrackstersRecoveryL1Seeded_cfi.py

@@ -0,0 +1,93 @@
+import FWCore.ParameterSet.Config as cms
+
+hltTiclTrackstersRecoveryL1Seeded = cms.EDProducer("TrackstersProducer",
+    detector = cms.string('HGCAL'),
+    filtered_mask = cms.InputTag("hltFilteredLayerClustersRecoveryL1Seeded","Recovery"),
+    itername = cms.string('Recovery'),
+    layer_clusters = cms.InputTag("hltHgcalMergeLayerClustersL1Seeded"),
+    layer_clusters_hfnose_tiles = cms.InputTag("ticlLayerTileHFNose"),
+    layer_clusters_tiles = cms.InputTag("hltTiclLayerTileProducerL1Seeded"),
+    mightGet = cms.optional.untracked.vstring,
+    original_mask = cms.InputTag("hltTiclTrackstersCLUE3DHighL1Seeded"),
+    patternRecognitionBy = cms.string('Recovery'),
+    inferenceAlgo = cms.string('TracksterInferenceByDNN'),
+    pluginPatternRecognitionByCA = cms.PSet(
+        algo_verbosity = cms.int32(0),
+        computeLocalTime = cms.bool(True),
+        energy_em_over_total_threshold = cms.double(-1),
+        etaLimitIncreaseWindow = cms.double(2.1),
+        filter_on_categories = cms.vint32(0),
+        max_delta_time = cms.double(3),
+        max_longitudinal_sigmaPCA = cms.double(9999),
+        max_missing_layers_in_trackster = cms.int32(9999),
+        max_out_in_hops = cms.int32(10),
+        min_cos_pointing = cms.double(-1),
+        min_cos_theta = cms.double(0.915),
+        min_layers_per_trackster = cms.int32(10),
+        oneTracksterPerTrackSeed = cms.bool(False),
+        out_in_dfs = cms.bool(True),
+        pid_threshold = cms.double(0),
+        promoteEmptyRegionToTrackster = cms.bool(False),
+        root_doublet_max_distance_from_seed_squared = cms.double(9999),
+        shower_start_max_layer = cms.int32(9999),
+        siblings_maxRSquared = cms.vdouble(0.0006, 0.0006, 0.0006),
+        skip_layers = cms.int32(0),
+        type = cms.string('CA')
+    ),
+    pluginPatternRecognitionByCLUE3D = cms.PSet(
+        algo_verbosity = cms.int32(0),
+        computeLocalTime = cms.bool(True),
+        criticalDensity = cms.vdouble(4, 4, 4),
+        criticalEtaPhiDistance = cms.vdouble(0.025, 0.025, 0.025),
+        criticalSelfDensity = cms.vdouble(0.15, 0.15, 0.15),
+        criticalXYDistance = cms.vdouble(1.8, 1.8, 1.8),
+        criticalZDistanceLyr = cms.vint32(5, 5, 5),
+        cutHadProb = cms.double(0.5),
+        densityEtaPhiDistanceSqr = cms.vdouble(0.0008, 0.0008, 0.0008),
+        densityOnSameLayer = cms.bool(False),
+        densitySiblingLayers = cms.vint32(3, 3, 3),
+        densityXYDistanceSqr = cms.vdouble(3.24, 3.24, 3.24),
+        doPidCut = cms.bool(False),
+        kernelDensityFactor = cms.vdouble(0.2, 0.2, 0.2),
+        minNumLayerCluster = cms.vint32(2, 2, 2),
+        nearestHigherOnSameLayer = cms.bool(False),
+        outlierMultiplier = cms.vdouble(2, 2, 2),
+        rescaleDensityByZ = cms.bool(False),
+        type = cms.string('CLUE3D'),
+        useAbsoluteProjectiveScale = cms.bool(True),
+        useClusterDimensionXY = cms.bool(False)
+    ),
+    pluginPatternRecognitionByFastJet = cms.PSet(
+        algo_verbosity = cms.int32(0),
+        antikt_radius = cms.double(0.09),
+        computeLocalTime = cms.bool(True),
+        minNumLayerCluster = cms.int32(5),
+        type = cms.string('FastJet')
+    ),
+    pluginPatternRecognitionByRecovery = cms.PSet(
+        algo_verbosity = cms.int32(0),
+        type = cms.string('Recovery')
+    ),
+
+    pluginInferenceAlgoTracksterInferenceByDNN = cms.PSet(
+        algo_verbosity = cms.int32(0),
+	onnxPIDModelPath = cms.FileInPath('RecoHGCal/TICL/data/ticlv5/onnx_models/patternrecognition/id_v0.onnx'),
+        onnxEnergyModelPath = cms.FileInPath('RecoHGCal/TICL/data/ticlv5/onnx_models/patternrecognition/energy_v0.onnx'),
+        inputNames  = cms.vstring('input'),
+        output_en   = cms.vstring('enreg_output'),
+        output_id   = cms.vstring('pid_output'),
+        eid_min_cluster_energy = cms.double(1),
+        eid_n_layers = cms.int32(50),
+        eid_n_clusters = cms.int32(10),
+        doPID = cms.int32(0),
+        doRegression = cms.int32(0),
+        type = cms.string('TracksterInferenceByDNN')
+    ),
+    pluginInferenceAlgoTracksterInferenceByANN = cms.PSet(
+      algo_verbosity = cms.int32(0),
+      type = cms.string('TracksterInferenceByANN')
+
+    ),
+    seeding_regions = cms.InputTag("hltTiclSeedingL1"),
+    time_layerclusters = cms.InputTag("hltHgcalMergeLayerClustersL1Seeded","timeLayerCluster")
+)

HLTrigger/Configuration/python/HLT_75e33/psets/hltTiclTracksterLinksPSet_cfi.py

@@ -0,0 +1,50 @@
+import FWCore.ParameterSet.Config as cms
+
+hltTiclTracksterLinksPSet = cms.PSet(
+      cylinder_radius_sqr_split = cms.double(9),
+      proj_distance_split = cms.double(5),
+      track_time_quality_threshold = cms.double(0.5),
+      min_num_lcs = cms.uint32(15),
+      min_trackster_energy = cms.double(20),
+      pca_quality_th = cms.double(0.85),
+      dot_prod_th = cms.double(0.97),
+      deltaRxy = cms.double(4),
+      lower_boundary = cms.vdouble(
+        20,
+        10
+      ),
+      upper_boundary = cms.vdouble(
+        150,
+        100
+      ),
+      upper_distance_projective_sqr = cms.vdouble(
+        30,
+        60
+      ),
+      lower_distance_projective_sqr = cms.vdouble(
+        30,
+        60
+      ),
+      min_distance_z = cms.vdouble(
+        35,
+        35
+      ),
+      upper_distance_projective_sqr_closest_points = cms.vdouble(
+        5,
+        30
+      ),
+      lower_distance_projective_sqr_closest_points = cms.vdouble(
+        10,
+        50
+      ),
+      max_z_distance_closest_points = cms.vdouble(
+        35,
+        35
+      ),
+      cylinder_radius_sqr = cms.vdouble(
+        9,
+        15
+      ),
+      algo_verbosity = cms.int32(0),
+      type = cms.string('Skeletons')
+    )

HLTrigger/Configuration/python/HLT_75e33/sequences/HLTHgcalTiclPFClusteringForEgammaUnseededSequence_cfi.py

@@ -18,7 +18,7 @@
 from ..modules.hltHgcalSoARecHitsLayerClustersProducer_cfi import *
 from ..modules.hltHgcalSoALayerClustersProducer_cfi import *
 from ..modules.hltHgcalLayerClustersFromSoAProducer_cfi import *
-from ..modules.hltTiclTracksterLinksUnseeded_cfi import *
+from ..modules.hltTiclTracksterLinks_cfi import *
 
 _HgcalLocalRecoUnseededSequence = cms.Sequence(hltHgcalDigis+hltHGCalUncalibRecHit+hltHGCalRecHit+hltParticleFlowRecHitHGC+hltHgcalLayerClustersEE+hltHgcalLayerClustersHSci+hltHgcalLayerClustersHSi+hltHgcalMergeLayerClusters)
 _HgcalTICLPatternRecognitionUnseededSequence = cms.Sequence(hltFilteredLayerClustersCLUE3DHigh+hltTiclSeedingGlobal+hltTiclLayerTileProducer+hltTiclTrackstersCLUE3DHigh)
@@ -49,7 +49,7 @@
 )
 
 # Use EGammaSuperClusterProducer at HLT in ticl v5
-hltTiclTracksterLinksSuperclusteringDNNUnseeded = hltTiclTracksterLinksUnseeded.clone(
+hltTiclTracksterLinksSuperclusteringDNNUnseeded = hltTiclTracksterLinks.clone(
     linkingPSet = cms.PSet(
         type=cms.string("SuperClusteringDNN"),
         algo_verbosity=cms.int32(0),
@@ -59,7 +59,7 @@
     tracksters_collections = [cms.InputTag("hltTiclTrackstersCLUE3DHigh")], # to be changed to ticlTrackstersCLUE3DEM once separate CLUE3D iterations are introduced
 )
 
-hltTiclTracksterLinksSuperclusteringMustacheUnseeded = hltTiclTracksterLinksUnseeded.clone(
+hltTiclTracksterLinksSuperclusteringMustacheUnseeded = hltTiclTracksterLinks.clone(
     linkingPSet = cms.PSet(
         type=cms.string("SuperClusteringMustache"),
         algo_verbosity=cms.int32(0)

HLTrigger/Configuration/python/HLT_75e33/sequences/HLTTiclTrackstersRecoverySequenceL1Seeded_cfi.py

@@ -0,0 +1,6 @@
+import FWCore.ParameterSet.Config as cms
+
+from ..modules.hltFilteredLayerClustersRecoveryL1Seeded_cfi import *
+from ..modules.hltTiclTrackstersRecoveryL1Seeded_cfi import *
+
+HLTTiclTrackstersRecoverySequence = cms.Sequence(hltFilteredLayerClustersRecovery+hltTiclTrackstersRecovery)

RecoHGCal/TICL/plugins/TICLDumper.cc

@@ -595,10 +595,10 @@ class TICLDumper : public edm::one::EDAnalyzer<edm::one::WatchRuns, edm::one::Sh
   const edm::EDGetTokenT<std::vector<reco::Muon>> muons_token_;
   const edm::EDGetTokenT<edm::ValueMap<std::pair<float, float>>> clustersTime_token_;
   const edm::EDGetTokenT<std::vector<int>> tracksterSeeds_token_;
-  const edm::EDGetTokenT<std::vector<std::vector<unsigned int>>> superclustering_linkedResultTracksters_token;
-  const edm::EDGetTokenT<reco::SuperClusterCollection> recoSuperClusters_token;
-  const edm::EDGetTokenT<reco::CaloClusterCollection> recoSuperClusters_caloClusters_token;
-  const edm::EDGetTokenT<std::vector<ticl::Trackster>> recoSuperClusters_sourceTracksters_token;
+  edm::EDGetTokenT<std::vector<std::vector<unsigned int>>> superclustering_linkedResultTracksters_token;
+  edm::EDGetTokenT<reco::SuperClusterCollection> recoSuperClusters_token;
+  edm::EDGetTokenT<reco::CaloClusterCollection> recoSuperClusters_caloClusters_token;
+  edm::EDGetTokenT<std::vector<ticl::Trackster>> recoSuperClusters_sourceTracksters_token;
   edm::ESGetToken<CaloGeometry, CaloGeometryRecord> caloGeometry_token_;
   const edm::EDGetTokenT<std::vector<ticl::Trackster>> simTracksters_SC_token_;  // needed for simticlcandidate
   const edm::EDGetTokenT<std::vector<TICLCandidate>> simTICLCandidate_token_;
@@ -884,6 +884,16 @@ TICLDumper::TICLDumper(const edm::ParameterSet& ps)
       saveTICLCandidate_(ps.getParameter<bool>("saveSimTICLCandidate")),
       saveSimTICLCandidate_(ps.getParameter<bool>("saveSimTICLCandidate")),
       saveTracks_(ps.getParameter<bool>("saveTracks")) {
+  if (saveSuperclustering_) {
+    superclustering_linkedResultTracksters_token =
+        consumes<std::vector<std::vector<unsigned int>>>(ps.getParameter<edm::InputTag>("superclustering"));
+    recoSuperClusters_token =
+        consumes<reco::SuperClusterCollection>(ps.getParameter<edm::InputTag>("recoSuperClusters"));
+    recoSuperClusters_caloClusters_token =
+        consumes<reco::CaloClusterCollection>(ps.getParameter<edm::InputTag>("recoSuperClusters"));
+    recoSuperClusters_sourceTracksters_token = consumes<std::vector<ticl::Trackster>>(
+        ps.getParameter<edm::InputTag>("recoSuperClusters_sourceTracksterCollection"));
+  }
   std::string detectorName_ = (detector_ == "HFNose") ? "HGCalHFNoseSensitive" : "HGCalEESensitive";
   hdc_token_ =
       esConsumes<HGCalDDDConstants, IdealGeometryRecord, edm::Transition::BeginRun>(edm::ESInputTag("", detectorName_));