-
Notifications
You must be signed in to change notification settings - Fork 17
/
Copy pathvalidation.py
157 lines (142 loc) · 7.2 KB
/
validation.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
from torch.autograd import Variable
import warnings
warnings.filterwarnings("ignore")
import torch
import torch.optim as optim
from datetime import datetime, timedelta
from torch.utils.data import DataLoader
import torch.nn as nn
from data.dataset import NCPDataset, NCP2DDataset, NCPJPGDataset, NCPJPGDataset_new, NCPJPGtestDataset_new
import os, cv2
import numpy as np
def _validate(modelOutput, length, labels, total=None, wrong=None):
averageEnergies = torch.mean(modelOutput.data, 1)
for i in range(modelOutput.size(0)):
# print(modelOutput[i,:length[i]].sum(0).shape)
averageEnergies[i] = modelOutput[i, :length[i]].mean(0)
maxvalues, maxindices = torch.max(averageEnergies, 1)
# print(maxindices.cpu().numpy())
# print(labels.cpu().numpy())
count = 0
for i in range(0, labels.squeeze(1).size(0)):
l = int(labels.squeeze(1)[i].cpu())
if total is not None:
if l not in total:
total[l] = 1
else:
total[l] += 1
if maxindices[i] == labels.squeeze(1)[i]:
count += 1
else:
if wrong is not None:
if l not in wrong:
wrong[l] = 1
else:
wrong[l] += 1
return (averageEnergies, count)
class Validator():
def __init__(self, options, mode, model, savenpy):
self.R = 'R' in options['general'].keys()
self.model = model
self.cls_num = options['general']['class_num']
self.use_plus = options['general']['use_plus']
self.use_3d = options['general']['use_3d']
self.usecudnn = options["general"]["usecudnn"]
self.use_lstm = options["general"]["use_lstm"]
self.batchsize = options["input"]["batchsize"]
self.use_slice = options['general']['use_slice']
self.asinput = options['general']['plus_as_input']
self.USE_25D = options['general']['use25d']
if options['general']['use_slice']:
if self.USE_25D:
f = 'data/3cls_test.list'
self.validationdataset = NCPJPGtestDataset_new(options[mode]["data_root"],
options["training"]["padding"],
f, cls_num=self.cls_num, mod=options['general']['mod'],
options=options)
else:
self.validationdataset = NCPJPGDataset_new(options[mode]["data_root"],
options[mode]["index_root"],
options[mode]["padding"],
False, cls_num=self.cls_num,
mod=options['general']['mod'],
options=options)
else:
if options['general']['use_3d']:
self.validationdataset = NCPDataset(options[mode]["index_root"],
options[mode]["padding"],
False,
z_length=options["model"]["z_length"])
else:
self.validationdataset = NCP2DDataset(options[mode]["data_root"],
options[mode]["index_root"],
options[mode]["padding"],
False)
self.savingnpy = savenpy
self.tot_data = len(self.validationdataset)
self.validationdataloader = DataLoader(
self.validationdataset,
batch_size=options["input"]["batchsize"],
shuffle=True,
num_workers=options["input"]["numworkers"],
drop_last=False
)
self.mode = mode
self.epoch = 0
def __call__(self):
self.epoch += 1
with torch.no_grad():
print("Starting {}...".format(self.mode))
count = np.zeros((self.cls_num + self.use_plus * 2 * (1 - self.asinput)))
Matrix = np.zeros((self.cls_num, self.cls_num))
if self.use_3d:
validator_function = self.model.validator_function()
if self.use_lstm:
validator_function = _validate
self.model.eval()
LL = []
GG = []
AA = []
if (self.usecudnn):
net = nn.DataParallel(self.model).cuda()
error_dir = 'error/'
os.makedirs(error_dir, exist_ok=True)
cnt = 0
num_samples = np.zeros((self.cls_num + self.use_plus * 2 * (1 - self.asinput)))
for i_batch, sample_batched in enumerate(self.validationdataloader):
input = Variable(sample_batched['temporalvolume']).cuda()
labels = Variable(sample_batched['label']).cuda()
# length = len(sample_batched['length'][1])
names = sample_batched['name']
outputs, f = net(input)
output_numpy = np.exp(outputs.cpu().numpy())
output_numpy = output_numpy[:, [0, -1]]
# output_numpy_ab=np.sum(output_numpy[:,1:],1)
# output_numpy=np.stack([output_numpy[:,0],output_numpy_ab],-1)
output_numpy = output_numpy / output_numpy.sum(1, keepdims=True)
pre = np.array(output_numpy[:, -1] > 0.5, np.int)
isacc = labels.cpu().numpy().reshape(labels.size(0)) == pre
label_numpy = labels.cpu().numpy()[:, 0]
# argmax = (-vector.cpu().numpy()).argsort()
for i in range(labels.size(0)):
LL.append([names[i], output_numpy[i, -1], label_numpy[i]])
Matrix[label_numpy[i], pre[i]] += 1
if isacc[i] == 1:
count[labels[i]] += 1
num_samples[labels[i]] += 1
if self.mode == 'validation':
if labels[i] == 1 and output_numpy[i, -1] > 0.99:
I = np.array(input[i, :, :, :].cpu().numpy() * 255, np.uint8).transpose(1, 2, 0)[:, :,
[2, 1, 0]]
cv2.imwrite('/mnt/data9/covid_detector_jpgs/selected_train1/abnor/abnor_' +
names[i].split('/')[-1].split('.')[0] + '.jpg', I)
if labels[i] == 0 and output_numpy[i, 0] > 0.99:
I = np.array(input[i, :, :, :].cpu().numpy() * 255, np.uint8).transpose(1, 2, 0)[:, :,
[2, 1, 0]]
cv2.imwrite('/mnt/data9/covid_detector_jpgs/selected_train1/nor/nor_' +
names[i].split('/')[-1].split('.')[0] + '.jpg', I)
print(count[:self.cls_num].sum() / num_samples[:self.cls_num].sum(), np.mean(AA))
LL = np.array(LL)
np.save(self.savingnpy, LL)
print(Matrix)
return count / num_samples, count[:self.cls_num].sum() / num_samples[:self.cls_num].sum()