-
Notifications
You must be signed in to change notification settings - Fork 0
/
ParallelNets_test.py
146 lines (106 loc) · 4.71 KB
/
ParallelNets_test.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
"""
Tests for ParallelNets model for crack tip predictions.
1) Deviation/Accuracy: How far from the label is the prediction?
2) Reliability: How often does the network fail in providing any prediction?
3) Dice coefficient
Instructions: Needs trained UNet model and interim-data
"""
# Imports
import os
import numpy as np
import torch
from torchvision.transforms import Compose
from torch.utils.data import DataLoader
from src.dataprocessing.dataset import CrackTipDataset
from src.deep_learning import nets, loss
from src.dataprocessing import transforms
from src.deep_learning.evaluate import get_deviation
# Utility functions
def calculate_deviation(model, dataloader):
devi = []
for sample in dataloader:
inputs, targets = sample['input'], sample['target']
inputs = inputs.to(device)
outputs = model(inputs)
outputs = outputs[0].detach().to('cpu')
condition = torch.BoolTensor(targets == 2)
labels = torch.where(condition, 1, 0)
labels = labels.unsqueeze(1)
devi_i = get_deviation(outputs, labels)
devi.append(devi_i)
return np.concatenate(devi)
def calculate_reliability(model, dataloader, dataset):
unpredicted = 0
for sample in dataloader:
inputs = sample['input'].to(device)
outputs = model(inputs)
outputs = outputs[0].detach().to('cpu')
condition = torch.BoolTensor(outputs >= 0.5)
is_crack_tip = torch.where(condition, 1, 0)
for i in range(outputs.shape[0]):
prediction_i = torch.nonzero(is_crack_tip[i], as_tuple=False)[:, -2:] / 1.
if len(prediction_i) == 0:
unpredicted += 1
return 1. - unpredicted / len(dataset)
def calculate_dice(model, dataloader, criterion, dataset):
running_loss = 0.0
for sample in dataloader:
inputs = sample['input'].to(device)
targets = sample['target'].to('cpu')
condition = torch.BoolTensor(targets == 2)
labels = torch.where(condition, 1, 0)
labels = labels.unsqueeze(1)
outputs = model(inputs)
outputs = outputs[0].detach().to('cpu')
current_loss = criterion(outputs, labels)
running_loss += current_loss.item() * inputs.shape[0]
dice_loss = running_loss / len(dataset)
return 1. - dice_loss
# Load the model
MODEL_PATH = os.path.join('models')
MODEL_NAME = 'ParallelNets-1'
model = nets.ParallelNets(in_ch=2, out_ch=1, init_features=64, dropout_prob=0.2)
criterion = loss.DiceLoss()
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
model_path = os.path.join(MODEL_PATH, MODEL_NAME, MODEL_NAME + '.pt')
model.load_state_dict(torch.load(model_path, map_location=device))
model.to(device)
model.eval()
# Deviation and Dice coefficent
####################################################################################################
input_data_path = [
os.path.join('data', 'S_160_4.7', 'interim', 'lInputData_left.pt'),
os.path.join('data', 'S_160_4.7', 'interim', 'lInputData_right.pt')
]
label_data_path = [
os.path.join('data', 'S_160_4.7', 'interim', 'lGroundTruthData_left.pt'),
os.path.join('data', 'S_160_4.7', 'interim', 'lGroundTruthData_right.pt')
]
# Dataset and dataloader
trsfms = Compose([transforms.EnhanceTip(),
transforms.InputNormalization()])
dataset = CrackTipDataset(inputs=input_data_path, labels=label_data_path, transform=trsfms)
dataloader = DataLoader(dataset, batch_size=10, shuffle=False, num_workers=4)
# Calculate and print results
deviations_in_px = calculate_deviation(model, dataloader)
deviations_in_mm = deviations_in_px * 70 / 256
print(f'Deviation (mean) in mm: {np.mean(deviations_in_mm)}')
print(f'Deviation (std) in mm: {np.std(deviations_in_mm)}')
dice = calculate_dice(model, dataloader, criterion, dataset)
print(f'Dice coefficient: {dice}')
# Reliability
####################################################################################################
input_data_paths = [
os.path.join('data', 'S_160_4.7', 'interim', 'lInputData_left.pt'),
os.path.join('data', 'S_160_4.7', 'interim', 'lInputData_right.pt'),
os.path.join('data', 'S_160_2.0', 'interim', 'lInputData_left.pt'),
os.path.join('data', 'S_160_2.0', 'interim', 'lInputData_right.pt'),
os.path.join('data', 'S_950_1.6', 'interim', 'lInputData_left.pt'),
os.path.join('data', 'S_950_1.6', 'interim', 'lInputData_right.pt')
]
# Dataset and dataloader
trsfms = Compose([transforms.InputNormalization()])
dataset = CrackTipDataset(inputs=input_data_paths, transform=trsfms)
dataloader = DataLoader(dataset, batch_size=10, shuffle=False, num_workers=4)
reliability = calculate_reliability(model, dataloader, dataset)
print(f'Reliability in %: {reliability * 100}')