-
Notifications
You must be signed in to change notification settings - Fork 5
/
extract_features.py
174 lines (129 loc) · 6.67 KB
/
extract_features.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
from __future__ import print_function
import numpy as np
import argparse
import time
import os
import torch
import torchvision
import torchvision.transforms as transforms
from torch.autograd import Variable
import wideresnet_UoS as wrn
# Setting the name of the out-of-distribution dataset
# Tiny-ImageNet (crop): Imagenet
# Tiny-ImageNet (resize): Imagenet_resize
# LSUN (crop): LSUN
# LSUN (resize): LSUN_resize
parser = argparse.ArgumentParser(description='Pytorch Extracting Features from Neural Networks')
parser.add_argument('--path', default="WideResNet-CIFAR10", type=str, help='path to model')
parser.add_argument('--in_dataset', default="cifar10", type=str, help='in-distribution dataset')
parser.add_argument('--out_dataset', default="Imagenet_crop", type=str,
help='out-of-distribution dataset')
parser.add_argument('--droprate', default=0.3, type=float,help='dropout probability (default: 0.3)')
parser.add_argument('--mc', default=50, type=int,help='Number of Monte Carlo samples (default: 50)')
parser.add_argument('--no_in_dataset', dest='in_extract', action='store_false',
help='do not extract features for in-distribution dataset - Default=True')
parser.add_argument('--gpu', default=0, type=int, help='gpu index')
parser.set_defaults(in_extract=True)
start = time.time()
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((125.3 / 255, 123.0 / 255, 113.9 / 255), (63.0 / 255, 62.1 / 255.0, 66.7 / 255.0)),
])
def main():
args = parser.parse_args()
test(args.in_dataset, args.in_extract, args.out_dataset, args.path, args.droprate, args.mc, args.gpu)
def test(in_dataset, extract_in_features, dataName, modelPath, d_rate, MC_runs, CUDA_DEVICE):
if in_dataset == "cifar10":
net1 = wrn.WideResNet(28, 10, 10, d_rate)
if in_dataset == "cifar100":
net1 = wrn.WideResNet(28, 100, 10, d_rate)
checkpoint = torch.load("../models/{}/model_best.pth.tar".format(modelPath))
net1.load_state_dict(checkpoint['state_dict'])
net1.cuda(CUDA_DEVICE)
testsetout = torchvision.datasets.ImageFolder("../data/{}".format(dataName), transform=transform)
testloaderOut = torch.utils.data.DataLoader(testsetout, batch_size=1,
shuffle=False, num_workers=2)
if in_dataset == "cifar10":
testset = torchvision.datasets.CIFAR10(root='../data', train=False, download=True, transform=transform)
testloaderIn = torch.utils.data.DataLoader(testset, batch_size=1,
shuffle=False, num_workers=2)
trainset = torchvision.datasets.CIFAR10(root='../data', train=True, download=True, transform=transform)
trainloaderIn = torch.utils.data.DataLoader(trainset, batch_size=1,
shuffle=False, num_workers=2)
if in_dataset == "cifar100":
testset = torchvision.datasets.CIFAR100(root='../data', train=False, download=True, transform=transform)
testloaderIn = torch.utils.data.DataLoader(testset, batch_size=1,
shuffle=False, num_workers=2)
trainset = torchvision.datasets.CIFAR100(root='../data', train=True, download=True, transform=transform)
trainloaderIn = torch.utils.data.DataLoader(trainset, batch_size=1,
shuffle=False, num_workers=2)
testData(net1, CUDA_DEVICE, trainloaderIn, testloaderIn, testloaderOut, modelPath, dataName, extract_in_features,
MC_runs)
def testData(net1, CUDA_DEVICE, trainloader, testloader_in, testloader_out, pathSave, dataName, extract_in,
MC_runs):
savepath = "../features/" + pathSave + "/"
ensure_dir(savepath)
t0 = time.time()
net1.eval()
if extract_in:
print("Processing in-distribution images")
#######################################In-distribution###########################################
print("Processing in-distribution images: train")
features = []
labels = []
for j, data in enumerate(trainloader):
images, lab = data
inputs = Variable(images.cuda(CUDA_DEVICE))
labels_t = Variable(lab.cuda(CUDA_DEVICE))
outputs, feat = net1(inputs)
features.extend(list(feat.data.cpu().numpy()))
labels.extend(labels_t.data.cpu())
if j % 1000 == 0:
print("{:4}/{:4} batches processed, {:.1f} seconds used.".format(j+1, len(trainloader), time.time()-t0))
t0 = time.time()
np.save(savepath+'featuresTrain_in', np.array(features))
np.save(savepath+'labelsTrain_in', labels)
if hasattr(net1.fc, 'bias'):
if net1.fc.bias is not None:
np.save(savepath + 'bias', net1.fc.bias.cpu().detach().numpy())
t0 = time.time()
print("Processing in-distribution images: test")
features = []
for j, data in enumerate(testloader_in):
images, lab = data
feat_list = []
with torch.no_grad():
inputs = Variable(images.cuda(CUDA_DEVICE))
for mc in range(MC_runs):
feat_list.append(net1(inputs)[1][:, : , None])
feat_list = torch.cat(feat_list, dim=2)
features.append(feat_list)
if j % 1000 == 0:
print("{:4}/{:4} batches processed, {:.1f} seconds used.".format(j+1, len(testloader_in), time.time()-t0))
t0 = time.time()
features = torch.cat(features).cpu().detach().numpy()
np.save(savepath+'featuresTest_in', features)
t0 = time.time()
print("Processing out-of-distribution images")
###################################Out-of-Distributions#####################################
features = []
for j, data in enumerate(testloader_out):
images, _ = data
feat_list = []
with torch.no_grad():
inputs = Variable(images.cuda(CUDA_DEVICE))
for mc in range(MC_runs):
feat_list.append(net1(inputs)[1][:, : , None])
feat_list = torch.cat(feat_list, dim=2)
features.append(feat_list)
if j % 1000 == 0:
print("{:4}/{:4} batches processed, {:.1f} seconds used.".format(j+1, len(testloader_out), time.time()-t0))
t0 = time.time()
features = torch.cat(features).cpu().detach().numpy()
np.save(savepath+'features_out_'+dataName, features)
def ensure_dir(directory):
if not os.path.exists(directory):
os.makedirs(directory)
print(directory + " was created")
if __name__ == '__main__':
main()