-
Notifications
You must be signed in to change notification settings - Fork 0
/
FormatTrainingData.py
243 lines (209 loc) · 10.6 KB
/
FormatTrainingData.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
import torch
# import torchvision
# from torch.utils.tensorboard import SummaryWriter
from tensorboardX import SummaryWriter
# from torchvision import datasets, transforms
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import numpy as np
from MiscFunctions import get_loss_weights_v2, unflatten_pos, calc_logger_stats
from MiscFunctions import make_log_stat_dict, reravel_array, make_prediction_vector
from MiscFunctions import blockPrint, enablePrint
from FancyLoader import FancyLoader
from ModelFunctions import LSTMTagger, run_validation_pass
import random
import ROOT
from array import array
from larcv import larcv
import os
import argparse
def prepare_sequence(seq, to_ix):
idxs = [to_ix[w] for w in seq]
return torch.tensor(idxs, dtype=torch.long)
def prepare_sequence_steps(seq,long=False):
full_np = np.stack(seq,axis=0)
if not long:
return torch.tensor(full_np, dtype=torch.float)
else:
return torch.tensor(full_np, dtype=torch.long)
PARAMS = {}
PARAMS['USE_CONV_IM'] = True
PARAMS['LARMATCH_CKPT'] = '/home/jmills/workdir/TrackWalker/larmatch_ckpt/checkpoint.1974000th.tar'
# PARAMS['LARMATCH_CKPT'] = '/cluster/tufts/wongjiradlab/twongj01/ubdl/larflow/larmatchnet/grid_deploy_scripts/larmatch_kps_weights/checkpoint.1974000th.tar'
PARAMS['MASK_WC'] = False
PARAMS['MIN_TRACK_LENGTH'] = 3.0
PARAMS['HIDDEN_DIM'] =1024
PARAMS['PADDING'] =10
PARAMS['EMBEDDING_DIM'] =(PARAMS['PADDING']*2+1)*(PARAMS['PADDING']*2+1) # N_Features
if PARAMS['USE_CONV_IM']:
PARAMS['EMBEDDING_DIM'] = PARAMS['EMBEDDING_DIM']*16 # 16 Features per pixel in larmatch
PARAMS['NUM_CLASSES'] =(PARAMS['PADDING']*2+1)*(PARAMS['PADDING']*2+1)+1 # Bonus Class is for the end of track class
PARAMS['TRACKEND_CLASS'] = (PARAMS['PADDING']*2+1)**2
PARAMS['CENTERPOINT_ISEND'] = True
if PARAMS['CENTERPOINT_ISEND']:
PARAMS['NUM_CLASSES'] =(PARAMS['PADDING']*2+1)*(PARAMS['PADDING']*2+1) #No bonus end of track class
PARAMS['TRACKEND_CLASS'] = (PARAMS['NUM_CLASSES']-1)/2
# PARAMS['INFILE'] ="/home/jmills/workdir/TrackWalker/inputfiles/merged_dlreco_75e9707a-a05b-4cb7-a246-bedc2982ff7e.root"
PARAMS['INFILE'] ="/cluster/tufts/wongjiradlab/jmills09/TrackerNet_InputData/midlevel_files/small_00002_merged_dlreco_497be540-00f9-49a8-9f80-7846143c4fce.root"
PARAMS['TRACK_IDX'] =0
PARAMS['EVENT_IDX'] =0
PARAMS['ALWAYS_EDGE'] =True # True points are always placed at the edge of the Padded Box
# TENSORDIR'] ="runs/Pad20_Hidden1024_500Entries"
PARAMS['CLASSIFIER_NOT_DISTANCESHIFTER'] =True # True -> Predict Output Pixel to step to, False -> Predict X,Y shift to next point
PARAMS['NDIMENSIONS'] = 2 #Not configured to have 3 yet.
PARAMS['LEARNING_RATE'] =0.0001 # 0.01 is good for the classifier mode,
PARAMS['DO_TENSORLOG'] = True
PARAMS['TENSORDIR'] = None # Default runs/DATE_TIME
PARAMS['SAVE_MODEL'] = False #should the network save the model?
PARAMS['CHECKPOINT_EVERY_N_EPOCHS'] =10000 # if not saving then this doesn't matter
PARAMS['EPOCHS'] = 5
PARAMS['VALIDATION_EPOCH_LOGINTERVAL'] = 1
PARAMS['VALIDATION_TRACKIDX_LOGINTERVAL'] = 100
PARAMS['TRAIN_EPOCH_LOGINTERVAL'] = 1
PARAMS['TRAIN_TRACKIDX_LOGINTERVAL'] = 100
# PARAMS['DEVICE'] = 'cuda:0'
PARAMS['DEVICE'] = 'cpu'
PARAMS['LOAD_SIZE'] = 50
PARAMS['TRAIN_EPOCH_SIZE'] = 500
PARAMS['VAL_EPOCH_SIZE'] = int(0.8*PARAMS['TRAIN_EPOCH_SIZE'])
PARAMS['SHUFFLE_DATASET'] = False
PARAMS['VAL_IS_TRAIN'] = False # This will set the validation set equal to the training set
PARAMS['AREA_TARGET'] = True # Change network to be predicting
PARAMS['TARGET_BUFFER'] = 2
def parse_args():
"""Parse input arguments"""
parser = argparse.ArgumentParser(description='Train a X-RCNN network')
parser.add_argument(
'--infile', dest='infile', required=False,
help='infile')
parser.add_argument(
'--infileidx', dest='infileidx', required=False,
help='infilinfileidxeidx')
parser.add_argument(
'--folderidx', dest='folderidx', required=False,
help='folderidx')
parser.add_argument(
'--outdir', dest='outdir', required=False,
help='outdir')
return parser.parse_args()
def main():
# test with:
# python FormatTrainingData_Complex.py --infileidx 0 --outdir test/ --folderidx 0 --infile /cluster/tufts/wongjiradlab/larbys/data/mcc9/mcc9_v29e_dl_run3b_bnb_nu_overlay_nocrtremerge/data/00/01/71/93/merged_dlreco_b1379c2b-cd47-4140-a448-3a27e8f28fb8.root
print("Let's Get Started.")
args = parse_args()
print('Called with args:')
args.infile = "inputfiles/VAL_BnBOverLay_DLReco_Tracker.root"
args.outdir = "sizetest/"
args.folderidx = str(0)
args.infileidx = "2d"#str()
print(args)
PARAMS['INFILE'] =args.infile
nbins = PARAMS['PADDING']*2+1
# entry_per_file = 9999
entry_per_file = 9999
Loader = FancyLoader(PARAMS)
# iter = int(tot_entries/entry_per_file)+1
iter = 1
for i in range(0,iter):
# start_file = i*entry_per_file
# end_file = i*entry_per_file+entry_per_file
# if end_file > tot_entries:
# end_file = tot_entries-1
start_entry = 0
end_entry = 2
max_tracks = 2
larmatch_feature_image_v, wire_image_v, step_idx_image_v, run_v, \
subrun_v, eventid_v, entry_v, mctrack_idx_v, mctrack_length_v, \
mctrack_pdg_v, mctrack_energy_v, charge_in_wire_v, charge_in_truth_v \
= Loader.load_fancy(start_entry, end_entry)
# larmatch_feature_image_v, wire_image_v, step_idx_image_v, run_v, subrun_v, eventid_v, entry_v, mctrack_idx_v, mctrack_length_v, mctrack_pdg_v, mctrack_energy_v, charge_in_wire_v, charge_in_truth_v = Loader.load_fancy()
# from MiscFunctions import save_im
# save_im(wire_image_v[0], savename="test/wireimg",canv_x=4000,canv_y=1000)
# save_im(step_idx_image_v[0], savename="test/stepimg",canv_x=4000,canv_y=1000)
outfile_str = "Wire_" if PARAMS['USE_CONV_IM'] == False else "LArMatch_"
outfile_str = outfile_str + "ComplexTrackIdx_"
outfile_str = args.outdir+args.folderidx+"/Reformat_" + outfile_str
if not os.path.exists(args.outdir+args.folderidx):
os.makedirs(args.outdir+args.folderidx)
outfile = ROOT.TFile(outfile_str+str(args.infileidx).zfill(3)+".root","recreate")
tree = ROOT.TTree("TrackWalkerInput_ComplexTrackIdx","TrackWalker Reformmated Input Tree Special Track Index Image")
wire_image_np = larcv.NumpyArrayFloat()
tree.Branch("wire_image_np",wire_image_np)
larmatchfeat_image_np = larcv.NumpyArrayFloat()
tree.Branch("larmatchfeat_image_np",larmatchfeat_image_np)
# stepidx_image_np = larcv.NumpyArrayFloat()
# tree.Branch("stepidx_image_np",stepidx_image_np)
#
# entry = array('f',[0]) # single float
# tree.Branch("original_entry",entry,"original_entry/F")
# mctrack_idx = array('f',[0]) # single float
# tree.Branch("mctrack_idx",mctrack_idx,"mctrack_idx/F")
# mctrack_length = array('f',[0]) # single float
# tree.Branch("mctrack_length",mctrack_length,"mctrack_length/F")
# mctrack_pdg = array('f',[0]) # single float
# tree.Branch("mctrack_pdg",mctrack_pdg,"mctrack_pdg/F")
# mctrack_energy = array('f',[0]) # single float
# tree.Branch("mctrack_energy",mctrack_energy,"mctrack_energy/F")
# run = array('f',[0]) # single float
# tree.Branch("run",run,"run/F")
# subrun = array('f',[0]) # single float
# tree.Branch("subrun",subrun,"subrun/F")
# event_id = array('f',[0]) # single float
# tree.Branch("event_id",event_id,"event_id/F")
#
# charge_in_wire = array('f',[0]) # single float
# tree.Branch("charge_in_wire",charge_in_wire,"charge_in_wire/F")
# charge_in_truth = array('f',[0]) # single float
# tree.Branch("charge_in_truth",charge_in_truth,"charge_in_truth/F")
print("Number of Tracks:", len(wire_image_v))
for idxx in range(len(wire_image_v)):
if idxx == 0:
continue
print("Filling Tree", idxx)
print("")
print(" ",wire_image_v[idxx].shape, "Wire Image Shape")
nFeats = wire_image_v[idxx].shape[0] * wire_image_v[idxx].shape[1]
print(" ",nFeats, "Wire Image Feats")
print(" ",larmatch_feature_image_v[idxx].shape, "LArFeat Image Shape")
nFeats = larmatch_feature_image_v[idxx].shape[0] * larmatch_feature_image_v[idxx].shape[1] * larmatch_feature_image_v[idxx].shape[2]
print(" ",nFeats, "LArFeat Image Feats")
nonzeroIdx = np.nonzero(larmatch_feature_image_v[idxx])
nFeats = nonzeroIdx[0].shape[0]
sFeats = np.zeros((nFeats,4)) #x,y,nfeat,val
for iii in range(nFeats):
sFeats[iii,0] = nonzeroIdx[0][iii]
sFeats[iii,1] = nonzeroIdx[1][iii]
sFeats[iii,2] = nonzeroIdx[2][iii]
sFeats[iii,3] = larmatch_feature_image_v[idxx][nonzeroIdx[0][iii],nonzeroIdx[1][iii],nonzeroIdx[2][iii]]
print("Storing sparselar")
print(wire_image_v[idxx].shape)
print(sFeats.shape)
wire_image_np.store(wire_image_v[idxx].astype(np.float32))
larmatchfeat_image_np.store(sFeats.astype(np.float32))
# print("Storing")
# print(wire_image_v[idxx].shape)
# print(larmatch_feature_image_v[idxx].shape)
# wire_image_np.store(wire_image_v[idxx].astype(np.float32))
# larmatchfeat_image_np.store(larmatch_feature_image_v[idxx].astype(np.float32))
# stepidx_image_np.store(step_idx_image_v[idxx].astype(np.float32))
#
# entry[0] = entry_v[idxx]
# mctrack_idx[0] = mctrack_idx_v[idxx]
# mctrack_length[0] = mctrack_length_v[idxx]
# mctrack_pdg[0] = mctrack_pdg_v[idxx]
# mctrack_energy[0] = mctrack_energy_v[idxx]
#
# run[0] = run_v[idxx]
# subrun[0] = subrun_v[idxx]
# event_id[0] = eventid_v[idxx]
#
# charge_in_wire[0] = charge_in_wire_v[idxx]
# charge_in_truth[0] = charge_in_truth_v[idxx]
tree.Fill()
idxx += 1
tree.Write()
outfile.Close()
return 0
if __name__ == '__main__':
main()