-
Notifications
You must be signed in to change notification settings - Fork 0
/
analyses.py
254 lines (205 loc) · 9 KB
/
analyses.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
import pandas as pd
import torch
from torch.optim import SGD, Adam
from torch.utils.data import DataLoader
from models import ViewMriNet, MainMriNet, load_checkpoint
from transforms import test_transforms
from view_model_training import ViewDataset
from main_model_training import MainDataset
import logging
from torch.nn.functional import softmax
from sklearn.metrics import roc_auc_score
def validate_model(
checkpoint_path: str,
root_dir: str,
device,
fill_observation_report: bool,
abnormality_type: str,
view_type: str,
pretrained_model_type: str
):
'''
- TP, TN, FP TN
- precission, recall, f1 score
- info which observation was properly classified
'''
# observations, preds, labels
stats = {
"abnormality_type": [abnormality_type],
"view_type": [view_type],
"pretrained_model_type": [pretrained_model_type]
}
ids = []
preds = []
labels_list = []
with torch.no_grad():
# model, transfer learning type irrelevant
model = ViewMriNet(pretrained_model_type, "fine_tunning")
# optimizer = SGD(model.parameters(), lr=0.01)
optimizer = Adam(model.parameters(), lr=1e-5)
model, optimizer, last_epoch = load_checkpoint(model, optimizer, checkpoint_path)
if torch.cuda.is_available():
model = model.to(device)
model.eval()
for state in ["train", "test"]:
logging.info(f"Started validation for {state}")
# calculated parameters
running_tp = 0
running_fp = 0
running_tn = 0
running_fn = 0
y_pred = []
y = []
# dataset, dataloader
dataset = ViewDataset(root_dir, state, view_type, abnormality_type, transform = test_transforms)
dataloader = DataLoader(dataset, batch_size=1)
len_dataset = len(dataset)
for id, batch in enumerate(dataloader, 0):
# progress
if id % 100 == 0 and id != 0:
progress = round(((id + 1) / len_dataset) * 100, 1)
logging.info(f"Progress: {progress}%")
# send images, labels to device
images, labels = batch
labels = labels[0]
y.append(int(labels.tolist()[1]))
if torch.cuda.is_available():
labels = labels.to(device)
images = images.to(device)
# calculate loss
outputs = model(images)
if torch.cuda.is_available():
outputs = outputs.to(device)
proba = softmax(outputs)
pred = torch.round(proba)
y_pred.append(int(pred.tolist()[1]))
# tp, fp, tn, fn
if torch.all(torch.eq(pred, labels)):
if pred.tolist() == [0, 1]:
# print("running_tp += 1")
running_tp += 1
else:
# print("running_tn += 1")
running_tn += 1
else:
if pred.tolist() == [1, 0]:
# print("running_fn += 1")
running_fn += 1
else:
# print("running_fp += 1")
running_fp += 1
if state == "test" and fill_observation_report:
ids.append(id)
preds.append(pred.tolist()[1])
labels_list.append(labels.tolist()[1])
# statistics
accuracy = round((running_tp + running_tn) / len_dataset, 2)
precission = round(running_tp / (running_tp + running_fp), 2) if running_tp + running_fp else 0
recall = round(running_tp / (running_tp + running_fn), 2) if running_tp + running_fn else 0
f1_score = round((2 * precission * recall) / (precission + recall), 2) if precission + recall else 0
roc_auc = roc_auc_score(y, y_pred)
stats[f"{state}_accuracy"] = [accuracy]
stats[f"{state}_precission"] = [precission]
stats[f"{state}_recall"] = [recall]
stats[f"{state}_f1_score"] = [f1_score]
stats[f"{state}_roc_auc"] = [roc_auc]
stats["epoch"] = [last_epoch]
stats = pd.DataFrame(stats)
# predictions for concrete observations (only for test)
observations_report = pd.DataFrame({"id": ids, "preds": preds, "labels": labels_list})
observations_report["pretrained_model_type"] = pretrained_model_type
observations_report["epoch"] = last_epoch
observations_report["view_type"] = view_type
observations_report["abnormality_type"] = abnormality_type
return stats, observations_report
def validate_main_model(
model_path: str,
checkpoint_path: str,
root_dir: str,
device,
fill_observation_report: bool,
abnormality_type: str
):
'''
- TP, TN, FP TN
- precission, recall, f1 score
- info which observation was properly classified
'''
# observations, preds, labels
stats = {"abnormality_type": [abnormality_type]}
ids = []
preds = []
labels_list = []
with torch.no_grad():
# model, transfer learning type irrelevant
model = MainMriNet(model_path, abnormality_type, "fine_tunning")
optimizer = Adam(model.final_classifier.parameters(), lr=1e-5)
model, optimizer, last_epoch = load_checkpoint(model, optimizer, checkpoint_path)
if torch.cuda.is_available():
model = model.to(device)
model.eval()
for state in ["train", "test"]:
logging.info(f"Started validation for {state}")
# calculated parameters
running_tp = 0
running_fp = 0
running_tn = 0
running_fn = 0
y_pred = []
y = []
# dataset, dataloader
dataset = MainDataset(root_dir, state, abnormality_type, transform = test_transforms)
dataloader = DataLoader(dataset, batch_size=1)
len_dataset = len(dataset)
for id, batch in enumerate(dataloader, 0):
# progress
if id % 100 == 0 and id != 0:
progress = round(((id + 1) / len_dataset) * 100, 1)
logging.info(f"Progress: {progress}%")
# send images, labels to device
image_axial, image_coronal, image_sagittal, labels = batch
image_axial = image_axial.to(device)
image_coronal = image_coronal.to(device)
image_sagittal = image_sagittal.to(device)
labels = labels[0].to(device)
y.append(int(labels.tolist()[1]))
# calculate loss
outputs = model(image_axial, image_coronal, image_sagittal).to(device)
if torch.cuda.is_available():
outputs = outputs.to(device)
proba = softmax(outputs)
pred = torch.round(proba)
y_pred.append(int(pred.tolist()[1]))
# tp, fp, tn, fn
if torch.all(torch.eq(pred, labels)):
if pred.tolist() == [0, 1]:
running_tp += 1
else:
running_tn += 1
else:
if pred.tolist() == [1, 0]:
running_fn += 1
else:
running_fp += 1
if state == "test" and fill_observation_report:
ids.append(id)
preds.append(pred.tolist()[1])
labels_list.append(labels.tolist()[1])
# statistics
accuracy = round((running_tp + running_tn) / len_dataset, 2)
precission = round(running_tp / (running_tp + running_fp), 2) if running_tp + running_fp else 0
recall = round(running_tp / (running_tp + running_fn), 2) if running_tp + running_fn else 0
f1_score = round((2 * precission * recall) / (precission + recall), 2) if precission + recall else 0
roc_auc = roc_auc_score(y, y_pred)
stats[f"{state}_accuracy"] = [accuracy]
stats[f"{state}_precission"] = [precission]
stats[f"{state}_recall"] = [recall]
stats[f"{state}_f1_score"] = [f1_score]
stats[f"{state}_roc_auc"] = [roc_auc]
stats["epoch"] = [last_epoch]
stats = pd.DataFrame(stats)
# predictions for concrete observations (only for test)
observations_report = pd.DataFrame({"id": ids, "preds": preds, "labels": labels_list})
observations_report["epoch"] = last_epoch
observations_report["abnormality_type"] = abnormality_type
return stats, observations_report