-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathutils.py
236 lines (221 loc) · 13.1 KB
/
utils.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
import torchvision
import torch.nn as nn
import torch
import timm
import torchvision.transforms as transforms
from torchvision.transforms import InterpolationMode
import PyTorch_CIFAR10.cifar10_models.vgg as cifar10_vgg
import PyTorch_CIFAR10.cifar10_models.resnet as cifar10_resnet
import PyTorch_CIFAR10.cifar10_models.mobilenetv2 as cifar10_mobilenetv2
import PyTorch_CIFAR10.cifar10_models.inception as cifar10_inception
import PyTorch_CIFAR10.cifar10_models.densenet as cifar10_densenet
import PyTorch_CIFAR10.cifar10_models.googlenet as cifar10_googlenet
import DVERGE.utils
import DVERGE.eval.eval_bbox
from MadryLab import model_utils as Madry_model_utils
from MadryLab.datasets import ImageNet as Madry_ImageNet
import os
import torch.nn.functional as F
dir_path = os.path.dirname(os.path.realpath(__file__))
## simple wrapper model to normalize an input image
class WrapperModel(nn.Module):
def __init__(self, model, mean, std,resize=False):
super(WrapperModel, self).__init__()
self.mean = torch.Tensor(mean)
self.model=model
self.resize=resize
self.std = torch.Tensor(std)
def forward(self, x):
if self.resize==True:
x = transforms.Resize((224, 224), interpolation=InterpolationMode.NEAREST)(x)
return self.model((x - self.mean.type_as(x)[None,:,None,None]) / self.std.type_as(x)[None,:,None,None])
def normalize(self, x):
if self.resize==True:
x = transforms.Resize((224, 224), interpolation=InterpolationMode.NEAREST)(x)
return (x - self.mean.type_as(x)[None,:,None,None]) / self.std.type_as(x)[None,:,None,None]
def load_model(model_name):
if model_name == "ResNet101":
model = torchvision.models.resnet101(pretrained=True)
elif model_name == 'ResNet18':
model = torchvision.models.resnet18(pretrained=True)
elif model_name == 'ResNet34':
model = torchvision.models.resnet34(pretrained=True)
elif model_name == 'ResNet50':
model = torchvision.models.resnet50(pretrained=True)
elif model_name == "ResNet152":
model = torchvision.models.resnet152(pretrained=True)
elif model_name == "vgg16":
model = torchvision.models.vgg16_bn(pretrained=True)
elif model_name == "vgg19":
model = torchvision.models.vgg19_bn(pretrained=True)
elif model_name == "wide_resnet101_2":
model = torchvision.models.wide_resnet101_2(pretrained=True)
elif model_name == "inception_v3":
model = torchvision.models.inception_v3(pretrained=True,transform_input=True)
elif model_name == "resnext50_32x4d":
model = torchvision.models.resnext50_32x4d(pretrained=True)
elif model_name == "alexnet":
model = torchvision.models.alexnet(pretrained=True)
elif model_name == "mobilenet_v3_large":
model = torchvision.models.mobilenet.mobilenet_v3_large(pretrained=True)
elif model_name == 'DenseNet121':
model = torchvision.models.densenet121(pretrained=True)
elif model_name == "DenseNet161":
model = torchvision.models.densenet161(pretrained=True)
elif model_name == 'mobilenet_v2':
model = torchvision.models.mobilenet_v2(pretrained=True)
elif model_name == "shufflenet_v2_x1_0":
model = torchvision.models.shufflenet_v2_x1_0(pretrained=True)
elif model_name == 'GoogLeNet':
model = torchvision.models.googlenet(pretrained=True)
# timm models
elif model_name == "efficientnet_b0":
model = timm.create_model('efficientnet_b0', pretrained=True)
elif model_name == "inception_resnet_v2":
model = timm.create_model("inception_resnet_v2", pretrained=True)
elif model_name == "inception_v3_timm":
model = timm.create_model("inception_v3", pretrained=True)
elif model_name == "inception_v4_timm":
model = timm.create_model("inception_v4", pretrained=True)
elif model_name == "xception":
model = timm.create_model("xception", pretrained=True)
# https://github.com/microsoft/robust-models-transfer
elif model_name == "resnet18_l2_eps0_1":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet18', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet18_l2_eps0.1.ckpt")
model = m.model
elif model_name == "resnet18_l2_eps1":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet18', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet18_l2_eps1.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps0_1":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps0.1.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps1":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps1.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps0_01":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps0.01.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps0_03":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps0.03.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps0_05":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps0.05.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps0_25":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps0.25.ckpt")
model = m.model
elif model_name == "resnet50_l2_eps0_5":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_l2_eps0.5.ckpt")
model = m.model
elif model_name == "resnet50_linf_eps0_5":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_linf_eps0.5.ckpt")
model = m.model
elif model_name == "resnet50_linf_eps1_0":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_linf_eps1.0.ckpt")
model = m.model
elif model_name == "resnet50_linf_eps2_0":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_linf_eps2.0.ckpt")
model = m.model
elif model_name == "resnet50_linf_eps4_0":
m, _ = Madry_model_utils.make_and_restore_model(arch='resnet50', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/resnet50_linf_eps4.0.ckpt")
model = m.model
elif model_name == "wide_resnet50_2_l2_eps0_1":
m, _ = Madry_model_utils.make_and_restore_model(arch='wide_resnet50_2', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/wide_resnet50_2_l2_eps0.1.ckpt")
model = m.model
elif model_name == "wide_resnet50_2_l2_eps1":
m, _ = Madry_model_utils.make_and_restore_model(arch='wide_resnet50_2', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/wide_resnet50_2_l2_eps1.ckpt")
model = m.model
elif model_name == "wide_resnet50_4_l2_eps0_1":
m, _ = Madry_model_utils.make_and_restore_model(arch='wide_resnet50_4', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/wide_resnet50_4_l2_eps0.1.ckpt")
model = m.model
elif model_name == "wide_resnet50_4_l2_eps1":
m, _ = Madry_model_utils.make_and_restore_model(arch='wide_resnet50_4', dataset=Madry_ImageNet(''),
resume_path=dir_path + "/microsoft_robust_models_transfer/src/checkpoints/wide_resnet50_4_l2_eps1.ckpt")
model = m.model
# timm Transformer-based models
elif model_name == "vit_base_patch16_224":
model = timm.create_model("vit_base_patch16_224", pretrained=True)
elif model_name == "levit_384":
model = timm.create_model("levit_384", pretrained=True)
elif model_name == "convit_base":
model = timm.create_model("convit_base", pretrained=True)
elif model_name == "twins_svt_base":
model = timm.create_model("twins_svt_base", pretrained=True)
elif model_name == "pit":
model = timm.create_model('pit_s_224', pretrained=True)
else:
raise ValueError(f"Not supported model name. {model_name}")
return model
def load_model_cifar10(model_name):
# https://github.com/huyvnphan/PyTorch_CIFAR10
if model_name == "vgg11_bn":
model = cifar10_vgg.vgg11_bn(pretrained=True)
elif model_name == "vgg13_bn":
model = cifar10_vgg.vgg13_bn(pretrained=True)
elif model_name == "vgg16_bn":
model = cifar10_vgg.vgg16_bn(pretrained=True)
elif model_name == "vgg19_bn":
model = cifar10_vgg.vgg19_bn(pretrained=True)
elif model_name == "resnet18":
model = cifar10_resnet.resnet18(pretrained=True)
elif model_name == "resnet34":
model = cifar10_resnet.resnet34(pretrained=True)
elif model_name == "resnet50":
model = cifar10_resnet.resnet50(pretrained=True)
elif model_name == "mobilenet_v2":
model = cifar10_mobilenetv2.mobilenet_v2(pretrained=True)
elif model_name == "inception_v3":
model = cifar10_inception.inception_v3(pretrained=True)
elif model_name == "densenet121":
model = cifar10_densenet.densenet121(pretrained=True)
elif model_name == "densenet161":
model = cifar10_densenet.densenet161(pretrained=True)
elif model_name == "densenet169":
model = cifar10_densenet.densenet169(pretrained=True)
elif model_name == "googlenet":
model = cifar10_googlenet.googlenet(pretrained=True)
# https://github.com/zjysteven/DVERGE
elif model_name == "ens3_res20_dverge":
args = DVERGE.eval.eval_bbox.get_args()
args.model_file = dir_path + "/DVERGE/checkpoints/dverge/seed_0/3_ResNet20_eps_0.07/epoch_200.pth"
model = DVERGE.utils.get_models(args, train=False, as_ensemble=True, model_file=args.model_file, leaky_relu=False)
elif model_name == "ens3_res20_gal":
args = DVERGE.eval.eval_bbox.get_args()
args.model_file =dir_path + "/DVERGE/checkpoints/gal/seed_0/3_ResNet20/epoch_200.pth"
model = DVERGE.utils.get_models(args, train=False, as_ensemble=True, model_file=args.model_file, leaky_relu=True)
elif model_name == "ens3_res20_adp":
args = DVERGE.eval.eval_bbox.get_args()
args.model_file = dir_path + "/DVERGE/checkpoints/adp/seed_0/3_ResNet20/epoch_200.pth"
model = DVERGE.utils.get_models(args, train=False, as_ensemble=True, model_file=args.model_file, leaky_relu=False)
elif model_name == "ens3_res20_baseline":
args = DVERGE.eval.eval_bbox.get_args()
args.model_file = dir_path + "/DVERGE/checkpoints/baseline/seed_0/3_ResNet20/epoch_200.pth"
model = DVERGE.utils.get_models(args, train=False, as_ensemble=True, model_file=args.model_file, leaky_relu=False)
else:
raise ValueError(f"Not supported model name. {model_name}")
model.eval()
return model
def str2bool(v):
if isinstance(v, bool):
return v
if v.lower() in ('yes', 'true', 't', 'y', '1'):
return True
elif v.lower() in ('no', 'false', 'f', 'n', '0'):
return False
else: raise argparse.ArgumentTypeError('Boolean value expected.')