LArbys / LArCV

Liquid Argon Computer Vision
11 stars 9 forks source link

labels filling correctly? #35

Closed twongjirad closed 8 years ago

twongjirad commented 8 years ago

From the Filler:

  [INFO]  <SimpleFiller::fill_entry_data::L320> Label to Datum: 1
  [INFO]  <test::_batch_process_::L259> Processed good event: valid entry counter = 2
  [INFO]  <test::_batch_process_::L248> Processing entry: 8378
  [INFO]  <testIOManager::get_data::L494> Setting event id (5360,0,35)
  [INFO]  <SimpleFiller::fill_entry_data::L307>     Cosmic PdgCode=0
Vertex   (x, y, z, t) = (0,0,0,0)
Momentum (px, py, pz) = (0,0,0)
Inittial Energy = 0
Deposit  Energy = 0
# Bounding Box  = 0

Looking into the ROOT file:

root [1] partroi_tpc_tree->Show(8378)
======> EVENT:8378
 partroi_tpc_branch = (larcv::EventROI*)0x4e8f630
 _producer       = tpc
 _run            = 5607
 _subrun         = 73
 _event          = 3672
 _part_v         = (vector<larcv::ROI>*)0x4e8f680
 _part_v._index  = 0, 1, 2, 3, 4
 _part_v._shape  = 2, 1, 1, 1, 1
 _part_v._type   = 2, 6, 9, 8, 8
 _part_v._mcst_index = 65535, 0, 1, 2, 4
 _part_v._mct_index = 0, 0, 0, 0, 0

Labeled as cosmic, but definitely a neutrino image.

—— the code ——

// labels
    ROIType_t roi_type = kROICosmic;
    for(auto const& roi : roi_v) {
      if(roi.MCSTIndex() != kINVALID_USHORT) continue;
      roi_type = roi.Type();
      if(roi_type == kROIUnknown) roi_type = PDG2ROIType(roi.PdgCode());
      LARCV_INFO() << roi.dump() << std::endl;
      break;
    }

    // Convert type to class
    size_t caffe_class = _roitype_to_class[roi_type];

    if(caffe_class == kINVALID_SIZE) {
      LARCV_CRITICAL() << "ROIType_t " << roi_type << " is not among those defined for final set of class!" << std::endl;
      throw larbys();
    }

    _label = (float)(_roitype_to_class[roi_type]);

Diagnosis

In the loop, the first entry it sees is a cosmic ROI.

It then breaks, missing the MC ROIs behind it. This get's labeled as cosmic, but it's a neutrino.

twongjirad commented 8 years ago

Why doesn't this happen every time?

Example of correct label:

    [INFO]  <test::_batch_process_::L248> Processing entry: 1783
      [INFO]  <testIOManager::get_data::L494> Setting event id (5498,22,1118)
      [INFO]  <SimpleFiller::fill_entry_data::L307>     BNB PdgCode=14
    Vertex   (x, y, z, t) = (137.517,16.5527,204.815,3403.09)
    Momentum (px, py, pz) = (-0.467529,1.17446,1243.1)
    Inittial Energy = 1243.1
    Deposit  Energy = 1098.56
    # Bounding Box  = 3
     Plane 0 (rows,cols) = (2160,672) ... Left Top (618,7806) ... Right Bottom (1290,5646)
     Plane 1 (rows,cols) = (2160,832) ... Left Top (709,7806) ... Right Bottom (1541,5646)
      Plane 2 (rows,cols) = (2160,1472) ... Left Top (670,7806) ... Right Bottom (2142,5646)
      [INFO]  <SimpleFiller::fill_entry_data::L320> Label to Datum: 1

What's in the ROOT file:

root [2] partroi_tpc_tree->Show(1783)
======> EVENT:1783
 partroi_tpc_branch = (larcv::EventROI*)0x4e8f630
 _producer       = tpc
 _run            = 5637
 _subrun         = 72
 _event          = 3610
 _part_v         = (vector<larcv::ROI>*)0x4e8f680
 _part_v._index  = 0, 1, 2
 _part_v._shape  = 2, 1, 1
 _part_v._type   = 2, 6, 9
 _part_v._mcst_index = 65535, 0, 1
 _part_v._mct_index = 0, 0, 0
 _part_v._energy_deposit = 271.893, 181.092, 90.801
 _part_v._energy_init = 414.253, 286.751, 1029.07
 _part_v._pdg    = 14, 13, 2212
 _part_v._parent_pdg = 0, 13, 2212
 _part_v._trackid = 4294967295, 1, 2
 _part_v._parent_trackid = 4294967295, 1, 2
 _part_v._vtx._x = 121.65, 121.65, 121.65
twongjirad commented 8 years ago

Oh shit! The Run,subrun,event numbers are different...

Something is not right ... I checked I am filling from same file I am checking with ROOT...

twongjirad commented 8 years ago

some notes:

use_thread = false, non-random: seems ok. it says it is processing entry X, and the ROI dumped out matches.

use_thread = true, non-random: seems ok

use_thread = true; random: weirdness

use_thread = false; random: weirdness

i'm relying on cout with threads -- dangerous? but it is single thread filling sequentially?

twongjirad commented 8 years ago

there a double-shuffle?

In initializer of https://github.com/LArbys/LArCV/blob/master/core/Processor/ProcessDriver.cxx the entries are shuffled. if(_random_access) std::random_shuffle(_access_entry_v.begin(),_access_entry_v.end());

In https://github.com/LArbys/LArCV/blob/master/app/APICaffe/ThreadDatumFiller.cxx _batch_process

the entry to be drawn is randomized between (0,numentries in tree) This random draw, X, is stored.

But when it goes to process driver, this X is used to get entry Y = _access_entry_v[X].

X is stored in vector passed back to analysis.

So when training, correct label for an image is sent to net. but the wrong entry number for image-ROI pair is passed back to analysis script.

twongjirad commented 8 years ago

Fixed in https://github.com/LArbys/LArCV/pull/36