-
Notifications
You must be signed in to change notification settings - Fork 0
/
coarse_testing.py
192 lines (181 loc) · 7.11 KB
/
coarse_testing.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
import numpy as np
import os
import sys
import time
from utils import *
from model import *
data_path = sys.argv[1]
current_fold = int(sys.argv[2])
organ_number = int(sys.argv[3])
low_range = int(sys.argv[4])
high_range = int(sys.argv[5])
slice_threshold = float(sys.argv[6])
slice_thickness = int(sys.argv[7])
organ_ID = int(sys.argv[8])
plane = sys.argv[9]
GPU_ID = int(sys.argv[10])
learning_rate1 = float(sys.argv[11])
learning_rate_m1 = int(sys.argv[12])
learning_rate2 = float(sys.argv[13])
learning_rate_m2 = int(sys.argv[14])
crop_margin = int(sys.argv[15])
crop_prob = float(sys.argv[16])
crop_sample_batch = int(sys.argv[17])
snapshot_path = os.path.join(snapshot_path, 'SIJ_training_' + \
sys.argv[11] + 'x' + str(learning_rate_m1) + ',' + str(crop_margin))
result_path = os.path.join(result_path, 'coarse_testing_' + \
sys.argv[11] + 'x' + str(learning_rate_m1) + ',' + str(crop_margin))
epoch = 'e' + sys.argv[18] + sys.argv[19] + sys.argv[20] + sys.argv[21]
epoch_list = [epoch]
timestamp = sys.argv[22]
snapshot_name = snapshot_name_from_timestamp(snapshot_path, \
current_fold, plane, 'I', slice_thickness, organ_ID, timestamp)
if snapshot_name == '':
exit('Error: no valid snapshot directories are detected!')
snapshot_directory = os.path.join(snapshot_path, snapshot_name)
print('Snapshot directory: ' + snapshot_directory + ' .')
snapshot = [snapshot_directory]
print(str(len(snapshot)) + ' snapshots are to be evaluated.')
for t in range(len(snapshot)):
print(' Snapshot #' + str(t + 1) + ': ' + snapshot[t] + ' .')
result_name = snapshot_name
os.environ["CUDA_VISIBLE_DEVICES"]= str(GPU_ID)
volume_list = open(testing_set_filename(current_fold), 'r').read().splitlines()
while volume_list[len(volume_list) - 1] == '':
volume_list.pop()
DSC = np.zeros((len(snapshot), len(volume_list)))
result_directory = os.path.join(result_path, result_name, 'volumes')
if not os.path.exists(result_directory):
os.makedirs(result_directory)
result_file = os.path.join(result_path, result_name, 'results.txt')
output = open(result_file, 'w')
output.close()
for t in range(len(snapshot)):
output = open(result_file, 'a+')
output.write('Evaluating snapshot ' + str(epoch_list[t]) + ':\n')
output.close()
finished = True
for i in range(len(volume_list)):
volume_file = volume_filename_testing(result_directory, epoch_list[t], i)
if not os.path.isfile(volume_file):
finished = False
break
if not finished:
net = RSTN(crop_margin=crop_margin, crop_prob=crop_prob, \
crop_sample_batch=crop_sample_batch, TEST='C').cuda()
net.load_state_dict(torch.load(snapshot[t]))
net.eval()
for i in range(len(volume_list)):
start_time = time.time()
print('Testing ' + str(i + 1) + ' out of ' + str(len(volume_list)) + ' testcases, ' + \
str(t + 1) + ' out of ' + str(len(snapshot)) + ' snapshots.')
volume_file = volume_filename_testing(result_directory, epoch_list[t], i)
s = volume_list[i].split(' ')
label = np.load(s[2])
label = is_organ(label, organ_ID).astype(np.uint8)
if not os.path.isfile(volume_file):
image = np.load(s[1]).astype(np.float32)
np.minimum(np.maximum(image, low_range, image), high_range, image)
image -= low_range
image /= (high_range - low_range)
print(' Data loading is finished: ' + \
str(time.time() - start_time) + ' second(s) elapsed.')
pred = np.zeros(image.shape, dtype = np.float32)
minR = 0
if plane == 'X':
maxR = image.shape[0]
shape_ = (1, 3, image.shape[1], image.shape[2])
elif plane == 'Y':
maxR = image.shape[1]
shape_ = (1, 3, image.shape[0], image.shape[2])
elif plane == 'Z':
maxR = image.shape[2]
shape_ = (1, 3, image.shape[0], image.shape[1])
for j in range(minR, maxR):
if slice_thickness == 1:
sID = [j, j, j]
elif slice_thickness == 3:
sID = [max(minR, j - 1), j, min(maxR - 1, j + 1)]
if plane == 'X':
image_ = image[sID, :, :].astype(np.float32)
elif plane == 'Y':
image_ = image[:, sID, :].transpose(1, 0, 2).astype(np.float32)
elif plane == 'Z':
image_ = image[:, :, sID].transpose(2, 0, 1).astype(np.float32)
image_ = image_.reshape((1, 3, image_.shape[1], image_.shape[2]))
image_ = torch.from_numpy(image_).cuda().float()
#pdb.set_trace()
out = net(image_, 1).data.cpu().numpy()[0, :, :, :]
if slice_thickness == 1:
if plane == 'X':
pred[j, :, :] = out
elif plane == 'Y':
pred[:, j, :] = out
elif plane == 'Z':
pred[:, :, j] = out
elif slice_thickness == 3:
if plane == 'X':
if j == minR:
pred[j: j + 2, :, :] += out[1: 3, :, :]
elif j == maxR - 1:
pred[j - 1: j + 1, :, :] += out[0: 2, :, :]
else:
pred[j - 1: j + 2, :, :] += out[...]
elif plane == 'Y':
if j == minR:
pred[:, j: j + 2, :] += out[1: 3, :, :].transpose(1, 0, 2)
elif j == maxR - 1:
pred[:, j - 1: j + 1, :] += out[0: 2, :, :].transpose(1, 0, 2)
else:
pred[:, j - 1: j + 2, :] += out[...].transpose(1, 0, 2)
elif plane == 'Z':
if j == minR:
pred[:, :, j: j + 2] += out[1: 3, :, :].transpose(1, 2, 0)
elif j == maxR - 1:
pred[:, :, j - 1: j + 1] += out[0: 2, :, :].transpose(1, 2, 0)
else:
pred[:, :, j - 1: j + 2] += out[...].transpose(1, 2, 0)
if slice_thickness == 3:
if plane == 'X':
pred[minR, :, :] /= 2
pred[minR + 1: maxR - 1, :, :] /= 3
pred[maxR - 1, :, :] /= 2
elif plane == 'Y':
pred[:, minR, :] /= 2
pred[:, minR + 1: maxR - 1, :] /= 3
pred[:, maxR - 1, :] /= 2
elif plane == 'Z':
pred[:, :, minR] /= 2
pred[:, :, minR + 1: maxR - 1] /= 3
pred[:, :, maxR - 1] /= 2
print(' Testing is finished: ' + str(time.time() - start_time) + ' second(s) elapsed.')
pred = np.around(pred * 255).astype(np.uint8)
np.savez_compressed(volume_file, volume = pred)
print(' Data saving is finished: ' + \
str(time.time() - start_time) + ' second(s) elapsed.')
pred_temp = (pred >= 128)
else:
volume_data = np.load(volume_file)
pred = volume_data['volume'].astype(np.uint8)
print(' Testing result is loaded: ' + \
str(time.time() - start_time) + ' second(s) elapsed.')
pred_temp = (pred >= 128)
DSC[t, i], inter_sum, pred_sum, label_sum = DSC_computation(label, pred_temp)
print(' DSC = 2 * ' + str(inter_sum) + ' / (' + str(pred_sum) + \
' + ' + str(label_sum) + ') = ' + str(DSC[t, i]) + ' .')
output = open(result_file, 'a+')
output.write(' Testcase ' + str(i + 1) + ': DSC = 2 * ' + str(inter_sum) + ' / (' + \
str(pred_sum) + ' + ' + str(label_sum) + ') = ' + str(DSC[t, i]) + ' .\n')
output.close()
if pred_sum == 0 and label_sum == 0:
DSC[t, i] = 0
print(' DSC computation is finished: ' + \
str(time.time() - start_time) + ' second(s) elapsed.')
print('Snapshot ' + str(epoch_list[t]) + ': average DSC = ' + str(np.mean(DSC[t, :])) + ' .')
output = open(result_file, 'a+')
output.write('Snapshot ' + str(epoch_list[t]) + \
': average DSC = ' + str(np.mean(DSC[t, :])) + ' .\n')
output.close()
print('The testing process is finished.')
for t in range(len(snapshot)):
print(' Snapshot ' + str(epoch_list[t]) + ': average DSC = ' + str(np.mean(DSC[t, :])) + ' .')