-
Notifications
You must be signed in to change notification settings - Fork 0
/
Train_ThumbClassifier.py
441 lines (289 loc) · 19.7 KB
/
Train_ThumbClassifier.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
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
# -*- coding: utf-8 -*-
"""
Created on Fri Aug 20 10:32:34 2021
@author: Lars Hilgers
"""
import Utils as utils
import os
import torch
import random
import argparse
import torchvision
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import torch.nn as nn
import torch.optim as optim
from sklearn import preprocessing
import sklearn.metrics as metrics
from sklearn.model_selection import StratifiedKFold
from Utils import DatasetLoader, Initialize_model
from Train_Utils import Train_model, Validate_model
from Utils import CalculateTotalROC, MergeResultCSV, Summarize_Classic, ConfusionMatrix, ReadExperimentFile, CreateProjectFolder, CalculateAccuracy
##############################################################################
parser = argparse.ArgumentParser(description='Main Script to Run Training')
parser.add_argument('--ExpFile', type = str, default = r"C:\Users\lhilgers\sciebo\Lars\Experiments\Thumb_Classifier\Experiments\Classic\Thresholds\Test\CR07_FullTrain_Classic_Test.txt", help = 'Adress to the experiment File')
args = parser.parse_args()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)
##############################################################################
if __name__ == '__main__':
args = ReadExperimentFile(args, deploy=True)
args.projectFolder = CreateProjectFolder(args.project_name, args.ExpFile)
os.mkdir(args.projectFolder)
# Load Ground Truth Data:
print('\n*** LOAD THE DATASET FOR TRAINING ***\n')
thumb_data = pd.read_excel(args.thumbData)
thumb_data = thumb_data[thumb_data["CLASSIFICATION"].notna()]
images = list(thumb_data["PATH"])
labels = list(thumb_data["CLASSIFICATION"])
le = preprocessing.LabelEncoder()
labels = le.fit_transform(labels)
args.num_classes = len(set(labels))
args.labelsDict = dict(zip(le.classes_, range(len(le.classes_))))
args.target_labels = list(le.classes_)
reportFile = open(os.path.join(args.projectFolder,'Report.txt'), 'a', encoding="utf-8")
reportFile.write('**********************************************************************' + '\n')
reportFile.write(str(args))
reportFile.write('\n' + '**********************************************************************' + '\n')
Summarize_Classic(args, list(labels), reportFile)
args.result = os.path.join(args.projectFolder, 'RESULTS')
os.makedirs(args.result, exist_ok = True)
args.split_dir = os.path.join(args.projectFolder, 'SPLITS')
os.makedirs(args.split_dir, exist_ok = True)
if args.trainFull:
print('\n***THIS IS A FULL TRAINING!***\n')
reportFile.write('\n***THIS IS A FULL TRAINING!***\n')
train_index = list(range(0,len(labels)))
val_index = random.choices(train_index, k=int(len(train_index) * 0.05))
val_x = [images[i] for i in val_index]
val_y = [labels[i] for i in val_index]
train_index = [i for i in train_index if i not in val_index]
train_x = [images[i] for i in train_index]
train_y = [labels[i] for i in train_index]
train_df = pd.DataFrame(zip(train_x,train_y))
train_subset = pd.DataFrame()
for tissue in set(train_df[1]):
train_subset = train_subset.append(train_df[train_df[1]==tissue].sample(n))
train_x = list(train_subset[0])
train_y = list(train_subset[1])
df = pd.DataFrame(list(zip(train_x, train_y)), columns=['PATH', 'CLASSIFICATION'])
df.to_csv(os.path.join(args.split_dir, 'FULL_TRAIN_TRAIN_SET' + '.csv'), index=False)
df = pd.DataFrame(list(zip(val_x, val_y)), columns=['PATH', 'CLASSIFICATION'])
df.to_csv(os.path.join(args.split_dir, 'FULL_TRAIN_VALIDATION_SET' + '.csv'), index=False)
model_ft, input_size = Initialize_model(args.model_name, args.num_classes, use_pretrained=True)
params = {'batch_size': args.batch_size,
'shuffle': True,
'num_workers': 16,
'pin_memory': False}
train_set = DatasetLoader(train_x, train_y, transform = torchvision.transforms.ToTensor, target_patch_size = input_size)
train_generator = torch.utils.data.DataLoader(train_set, **params)
if 'val_index' in locals():
val_set = DatasetLoader(val_x, val_y, transform=torchvision.transforms.ToTensor,
target_patch_size=input_size)
val_generator = torch.utils.data.DataLoader(val_set, **params)
noOfLayers = 0
for name, child in model_ft.named_children():
noOfLayers += 1
cut = int(args.freezeRatio * noOfLayers)
ct = 0
for name, child in model_ft.named_children():
ct += 1
if ct < cut:
for name2, params in child.named_parameters():
params.requires_grad = False
model_ft = model_ft.to(device)
# Optimizer
print('\n*** INITIALIZE THE OPTIMIZER***\n', end=' ')
reportFile.write('\n*** INITIALIZE THE OPTIMIZER***\n')
optimizer = optim.Adam(filter(lambda p: p.requires_grad, model_ft.parameters()) , lr = args.lr, weight_decay = args.wd)
criterion = nn.CrossEntropyLoss()
# Start of training
print('\n*** START TRAINING ***', end=' ')
model, train_loss_history, train_acc_history, val_acc_history, val_loss_history = Train_model(model=model_ft,
trainLoaders=train_generator,
valLoaders=val_generator,
criterion=criterion,
optimizer=optimizer,
num_epochs=args.max_epochs,
patience=args.patience,
stop_epoch=args.stopEpoch, trainFull = True, resultFolder = args.result)
torch.save(model.state_dict(), os.path.join(args.projectFolder, 'RESULTS', 'MODEL_Full_Final'))
df = pd.DataFrame(list(zip(train_loss_history, train_acc_history)),
columns=['train_loss_history', 'train_acc_history'])
df.to_csv(os.path.join(args.result, 'TRAIN_HISTORY_full' + '.csv'))
reportFile.close()
else:
print('**********************************************************************')
print('START OF CROSS VALIDATION! THIS RUN USES {} % OF THE DATASET!'.format(100))
print('**********************************************************************')
reportFile.write('**********************************************************************\n')
reportFile.write('START OF CROSS VALIDATION\n')
reportFile.write('**********************************************************************\n')
folds = args.foldNumber
kf = StratifiedKFold(n_splits=folds, random_state=args.seed, shuffle=True)
kf.get_n_splits(images, labels)
fold = 1
for train_index, test_index in kf.split(images, labels):
images = np.array(images)
labels = np.array(labels)
test_x = images[test_index]
test_y = labels[test_index]
if not len(images)<50:
val_index = random.choices(train_index, k = int(len(train_index) * 0.05))
val_x = images[val_index]
val_y = labels[val_index]
train_index = [i for i in train_index if i not in val_index]
train_x = images[train_index]
train_y = labels[train_index]
#train_df = pd.DataFrame(zip(train_x,train_y))
#train_subset = pd.DataFrame()
#print(train_df)
#for tissue in set(train_df[1]):
# train_subset = train_subset.append(train_df[train_df[1]==tissue].sample(n))
#rain_subset = pd.DataFrame(train_subset)
#print(train_subset)
#train_x = list(train_subset[0])
#train_y = list(train_subset[1])
#print(train_x, train_y)
df = pd.DataFrame(list(zip(train_x, train_y)), columns =['PATH', 'CLASSIFICATION'])
df.to_csv(os.path.join(args.split_dir, 'Train_' + str(fold)+ '.csv'), index = False)
df = pd.DataFrame(list(zip(test_x, test_y)), columns =['PATH', 'CLASSIFICATION'])
df.to_csv(os.path.join(args.split_dir, 'Test_' + str(fold)+ '.csv'), index = False)
df = pd.DataFrame(list(zip(val_x, val_y)), columns =['PATH', 'CLASSIFICATION'])
df.to_csv(os.path.join(args.split_dir, 'Validation_' + str(fold)+ '.csv'), index = False)
model_ft, input_size = Initialize_model(args.model_name, args.num_classes, use_pretrained = True)
if args.SSL:
model_ft.load_state_dict(torch.load(args.modelAdr),strict = False)
params = {'batch_size': args.batch_size,
'shuffle': True,
'num_workers': 16,
'pin_memory' : False}
train_set = DatasetLoader(train_x, train_y, transform = torchvision.transforms.ToTensor, target_patch_size = input_size)
train_generator = torch.utils.data.DataLoader(train_set, **params)
if 'val_index' in locals():
val_set = DatasetLoader(val_x, val_y, transform = torchvision.transforms.ToTensor, target_patch_size = input_size)
val_generator = torch.utils.data.DataLoader(val_set, **params)
params = {'batch_size': args.batch_size,
'shuffle': False,
'num_workers': 16,
'pin_memory' : False}
test_set = DatasetLoader(test_x, test_y, transform = torchvision.transforms.ToTensor, target_patch_size = input_size)
test_generator = torch.utils.data.DataLoader(test_set, **params)
noOfLayers = 0
for name, child in model_ft.named_children():
noOfLayers += 1
cut = int(args.freezeRatio * noOfLayers)
ct = 0
for name, child in model_ft.named_children():
ct += 1
if ct < cut:
for name2, params in child.named_parameters():
params.requires_grad = False
model_ft = model_ft.to(device)
optimizer = optim.Adam(filter(lambda p: p.requires_grad, model_ft.parameters()) , lr = args.lr, weight_decay = args.wd)
criterion = nn.CrossEntropyLoss()
model, train_loss_history, train_acc_history, val_acc_history, val_loss_history = Train_model(model = model_ft, trainLoaders = train_generator,valLoaders = val_generator, criterion = criterion,
optimizer = optimizer, num_epochs = args.max_epochs, patience = args.patience, stop_epoch = args.stopEpoch,
trainFull = False, resultFolder = args.result, fold = fold)
torch.save(model.state_dict(), os.path.join(args.result,'MODEL_FOLD_FINAL' + str(fold)))
df = pd.DataFrame(list(zip(train_loss_history, train_acc_history, val_loss_history, val_acc_history)),
columns =['train_loss_history', 'train_acc_history', 'val_loss_history', 'val_acc_history'])
df.to_csv(os.path.join(args.result, 'TRAIN_HISTORY_FOLD_' + str(fold) + '.csv'), index = False)
epoch_loss, epoch_acc, predList = Validate_model(model, test_generator, criterion, args.hasLabels)
predictions = {}
for index, key in enumerate(list(args.labelsDict.keys())):
predictions[key] = []
for item in predList:
predictions[key].append(item[index])
predictions = pd.DataFrame.from_dict(predictions)
results_df = pd.DataFrame(list(zip(test_x, test_y)), columns =['PATH', 'CLASSIFICATION'])
results_df = pd.concat([results_df, predictions], axis=1)
results_df.to_csv(os.path.join(args.result, 'TEST_RESULT_SCORES_' + str(fold) + '.csv'), index = False)
keys = list(args.labelsDict.keys())
for index, key in enumerate(keys):
fpr, tpr, thresholds = metrics.roc_curve(test_y, results_df[key], pos_label = args.labelsDict[key])
print('AUC FOR TARGET {} IN THIS DATA SET IN FOLD {} IS: {} '.format(key, fold, metrics.auc(fpr, tpr)))
reportFile.write('AUC FOR TARGET {} IN THIS DATA SET IN FOLD {} IS: {} '.format(key, fold, metrics.auc(fpr, tpr)) + '\n')
fold += 1
reportFile.write('**********************************************************************' + '\n')
test_scores = []
for i in range(args.foldNumber):
test_scores.append('TEST_RESULT_SCORES_' + str(i+1) + '.csv')
graph_dir = (os.path.join(args.result, "Graphs"))
os.mkdir(graph_dir)
######################
results_df = []
for item in test_scores:
data = pd.read_csv(os.path.join(args.result, item))
results_df.append(data)
results_df = pd.concat(results_df)
results_df["model_prediction"] = results_df.iloc[:, 2:].idxmax(axis=1)
keys = list(args.labelsDict.keys())
true_y = results_df["CLASSIFICATION"]
returnList = []
for key in keys:
pred_y = results_df[str(key)]
fpr, tpr, thresholds = metrics.roc_curve(true_y, pred_y, pos_label=args.labelsDict[key])
print('AUC FOR TARGET {} IN THIS DATA SET IS: {} '.format(key, metrics.auc(fpr, tpr)))
# reportFile.write('AUC FOR TARGET {} IN THIS DATA SET IS: {}\n '.format(key, metrics.auc(fpr, tpr)))
returnList.append('AUC FOR TARGET {} IN THIS DATA SET IS: {}\n '.format(key, metrics.auc(fpr, tpr)))
auc_values = []
nsamples = 1000
rng = np.random.RandomState(666)
y_true = np.array(true_y)
y_pred = np.array(pred_y)
for i in range(nsamples):
indices = rng.randint(0, len(y_pred), len(y_pred))
if len(np.unique(y_pred[indices])) < 2 or np.sum(y_true[indices]) == 0:
continue
fpr, tpr, thresholds = metrics.roc_curve(y_true[indices], y_pred[indices],
pos_label=args.labelsDict[key])
auc_values.append(metrics.auc(fpr, tpr))
auc_values = np.array(auc_values)
auc_values.sort()
print('Lower Confidence Interval For Target {}: {}'.format(key, np.round(
auc_values[int(0.025 * len(auc_values))], 10)))
# reportFile.write('Lower Confidence Interval For Target {}: {}'.format(key, np.round(
# auc_values[int(0.025 * len(auc_values))], 3)))
returnList.append('Lower Confidence Interval For Target {}: {}'.format(key, np.round(
auc_values[int(0.025 * len(auc_values))], 10)))
print('Higher Confidence Interval For Target {} : {}'.format(key, np.round(
auc_values[int(0.975 * len(auc_values))], 10)))
# reportFile.write('Higher Confidence Interval For Target {} : {}'.format(key, np.round(
# auc_values[int(0.975 * len(auc_values))], 3)))
returnList.append('Higher Confidence Interval For Target {} : {} \n'.format(key, np.round(
auc_values[int(0.975 * len(auc_values))], 10)))
fig, ax = plt.subplots()
plt.title('Receiver Operating Characteristic', fontsize=15)
plt.plot(fpr, tpr, color="black")
plt.plot([0, 1], color="navy", ls="--")
plt.ylabel('TPR (sensitivity', fontsize=15)
plt.xlabel('FPR (1-specificity)', fontsize=15)
ax.tick_params(axis='both', which='major', labelsize=15)
ax.tick_params(axis='both', which='minor', labelsize=15)
ax.set_aspect(1.0 / ax.get_data_ratio(), adjustable='box')
plt.savefig(os.path.join(graph_dir, str(key) + ".jpg"))
plt.clf()
inv_labelsDict = {v: k for k, v in args.labelsDict.items()}
true_y = true_y.replace(inv_labelsDict)
results_df["CLASSIFICATION"] = true_y
results_df.to_csv(os.path.join(args.result, 'THUMB_CLASSIFICATION_RESULTS.csv'), index=False)
accuracy = utils.CalculateAccuracy(results_df)
scores = results_df.iloc[:,2:(len(results_df.columns)-1)]
true_y = results_df["CLASSIFICATION"]
auc_total, auc_lower, auc_upper = utils.CalculateTotalAUROC(true_y, scores, args.target_labels)
returnList.append("Macro-averaged AUC for this data set: {}".format(auc_total))
returnList.append("Lower Confidence Interval for AUC for this data set: {}".format(auc_lower))
returnList.append("Higher Confidence Interval for AUC for this data set: {}".format(auc_upper))
print("Total Accuracy is: {:.4f}".format(accuracy))
# reportFile.write("Total Accuracy is: {:.4f}".format(accuracy))
returnList.append("Total Accuracy is: {:.4f}".format(accuracy))
utils.ConfusionMatrix_new(results_df, graph_dir, args.cm_xlabels, args.cm_ylabels, args.cm_sortlabels)
##############################
for item in returnList:
reportFile.write(item + '\n')
reportFile.write('**********************************************************************' + '\n')
MergeResultCSV(args.result, test_scores, returnList)
reportFile.close()
###############################################################################