-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathanomaly_detection.py
172 lines (148 loc) · 6.52 KB
/
anomaly_detection.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
import os, sys
import numpy as np
import torch
import json
import argparse
import pickle
import torchvision
import torchvision.transforms as transforms
from singleVis.SingleVisualizationModel import VisModel
from singleVis.data import NormalDataProvider
from singleVis.projector import Projector, tfDVIProjector, TimeVisProjector
parser = argparse.ArgumentParser()
parser.add_argument('--dataset', type=str, choices=["cifar10","mnist","fmnist"])
parser.add_argument('--noise_rate', type=int, default=5, choices=[5,10,20])
parser.add_argument("--batch_size", type=int, default=500)
parser.add_argument('--method', type=str, choices=["tfDVI", "TimeVis"])
args = parser.parse_args()
dataset = args.dataset
noise_type = "symmetric"
noise_rate = args.noise_rate
batch_size = args.batch_size
VIS_METHOD = args.method
inject_OOD = {
"mnist":"cifar10",
"fmnist":"mnist",
"cifar10":"mnist"
}
# tensorflow
visible_device = "0,1,2,3"
os.environ["CUDA_VISIBLE_DEVICES"] = visible_device
CONTENT_PATH = "/home/xianglin/projects/DVI_data/noisy/{}/{}/{}/".format(noise_type, dataset, noise_rate)
sys.path.append(CONTENT_PATH)
with open(os.path.join(CONTENT_PATH, "config.json"), "r") as f:
config = json.load(f)
config = config[VIS_METHOD]
CLASSES = config["CLASSES"]
GPU_ID = config["GPU"]
EPOCH_START = config["EPOCH_START"]
EPOCH_END = config["EPOCH_END"]
EPOCH_PERIOD = config["EPOCH_PERIOD"]
# Training parameter (subject model)
TRAINING_PARAMETER = config["TRAINING"]
NET = TRAINING_PARAMETER["NET"]
LEN = TRAINING_PARAMETER["train_num"]
test_len = TRAINING_PARAMETER["test_num"]
# Training parameter (visualization model)
VISUALIZATION_PARAMETER = config["VISUALIZATION"]
ENCODER_DIMS = VISUALIZATION_PARAMETER["ENCODER_DIMS"]
DECODER_DIMS = VISUALIZATION_PARAMETER["DECODER_DIMS"]
VIS_MODEL_NAME = VISUALIZATION_PARAMETER["VIS_MODEL_NAME"]
TOTOAL_EPOCH = (EPOCH_END-EPOCH_START)//EPOCH_PERIOD + 1
# define hyperparameters
DEVICE = torch.device("cuda:{}".format(GPU_ID) if torch.cuda.is_available() else "cpu")
import Model.model as subject_model
# net = resnet18()
net = eval("subject_model.{}()".format(NET))
data_provider = NormalDataProvider(CONTENT_PATH, net, EPOCH_START, EPOCH_END, EPOCH_PERIOD, device=DEVICE, classes=CLASSES, epoch_name="Epoch", verbose=1)
if VIS_METHOD == "tfDVI":
# Define Projector
flag = "_temporal_id_withoutB"
projector = tfDVIProjector(CONTENT_PATH, flag=flag)
elif VIS_METHOD == "TimeVis":
model = VisModel(ENCODER_DIMS, DECODER_DIMS)
projector = TimeVisProjector(vis_model=model, content_path=CONTENT_PATH, vis_model_name=VIS_MODEL_NAME, device=DEVICE)
elif VIS_METHOD == "DeepDebugger":
model = VisModel(ENCODER_DIMS, DECODER_DIMS)
SEGMENTS = VISUALIZATION_PARAMETER["SEGMENTS"]
projector = Projector(vis_model=model, content_path=CONTENT_PATH, segments=SEGMENTS, device=DEVICE)
with open(os.path.join(CONTENT_PATH, '{}_sample_recommender.pkl'.format(VIS_METHOD)), 'rb') as f:
tm = pickle.load(f)
if inject_OOD[dataset] == "mnist":
train_loader = torch.utils.data.DataLoader(
torchvision.datasets.MNIST('experiments/data/', train=True, download=True,
transform=torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(
# (0.1307,), (0.3081,))
(0.5,), (0.5,))])),
batch_size=batch_size, shuffle=True)
elif inject_OOD[dataset] == "cifar10":
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
trainset = torchvision.datasets.CIFAR10(root='experiments/data', train=True,
download=True, transform=transform)
train_loader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True)
elif inject_OOD[dataset] == "fmnist":
train_loader = torch.utils.data.DataLoader(
torchvision.datasets.FashionMNIST('expriments/data', train=True, download=True,
transform=torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(
(0.5,), (0.5,))
])),
batch_size=batch_size, shuffle=True)
detect_anomaly = lambda scores: max(scores)>0.98
anomaly = 0
# true sample
selected = np.random.choice(test_len, batch_size, replace=False)
samples = np.zeros((TOTOAL_EPOCH, batch_size, 512))
for i in range(EPOCH_START, EPOCH_END+1, EPOCH_PERIOD):
e = (i-EPOCH_START)//EPOCH_PERIOD
samples[e] = data_provider.test_representation(i)[selected]
embeddings_2d = np.zeros((TOTOAL_EPOCH, batch_size, 2))
for e in range(1, TOTOAL_EPOCH+1, 1):
embeddings_2d[e-1] = projector.batch_project(e, samples[e-1])
embeddings_2d = np.transpose(embeddings_2d, [1,0,2])
for i in range(batch_size):
scores = tm.score_new_sample(embeddings_2d[i][-tm.period:])
anomaly = anomaly+ detect_anomaly(scores)
print(scores, detect_anomaly(scores))
print(anomaly/batch_size)
normal_rate = anomaly/batch_size
# OOD
anomaly = 0
for img, target in train_loader:
image = img.detach().cpu().numpy()
if dataset == "mnist":
image = image[:, :1, :28,:28]
if dataset == "cifar10":
image = np.pad(image, ((0,0),(1,1), (2, 2), (2, 2)), 'edge')
embedding_2d = np.zeros((TOTOAL_EPOCH, batch_size, 2))
for i in range(EPOCH_START, EPOCH_END+1, EPOCH_PERIOD):
e = int((i-EPOCH_START)/EPOCH_PERIOD)
repr = data_provider.feature_function(i)(torch.from_numpy(image).to(DEVICE))
embedding_2d[e] = projector.batch_project(i, repr.detach().cpu().numpy())
embedding_2d = np.transpose(embedding_2d, [1,0,2])
for i in range(batch_size):
scores = tm.score_new_sample(embedding_2d[i][-tm.period:])
anomaly = anomaly+ detect_anomaly(scores)
print(scores, detect_anomaly(scores))
break
print(anomaly/batch_size)
anomaly_rate = anomaly/batch_size
save_file = "/home/xianglin/projects/git_space/DLVisDebugger/anomaly_detection.txt"
line = [VIS_METHOD, dataset, str(float(normal_rate)), str(float(anomaly_rate))]
if not os.path.exists(save_file):
with open(save_file, 'w') as f:
for word in line:
f.write(word)
f.write("\t")
f.write('\n')
else:
with open(save_file, 'a') as f:
for word in line:
f.write(word)
f.write("\t")
f.write('\n')