-
Notifications
You must be signed in to change notification settings - Fork 0
/
poison_generation.py
189 lines (154 loc) · 6.97 KB
/
poison_generation.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
import argparse
import numpy as np
from tqdm import tqdm
import torch
import torch.nn as nn
import torch.nn.functional as F
from dataloader.ModelNetDataLoader40 import ModelNetDataLoader40
from dataloader.ModelNetDataLoader10 import ModelNetDataLoader10
from dataloader.ShapeNetDataLoader import PartNormalDataset
from torch.utils.data import DataLoader, TensorDataset
from utils.logging import Logging_str
from utils.utils import show_time, transform_time, set_seed
import math
import os
import sys
import time
import random
import importlib
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
ROOT_DIR = BASE_DIR
sys.path.append(os.path.join(ROOT_DIR, 'classifiers'))
def load_data(args, data_path):
"""Load the dataset from the given path"""
print('Start Loading Dataset...')
if args.dataset == 'ModelNet40':
DATASET = ModelNetDataLoader40(
root=data_path,
npoint=args.input_point_nums,
split='train',
normal_channel=False
)
elif args.dataset == 'ModelNet10':
DATASET = ModelNetDataLoader10(
root=data_path,
npoint=args.input_point_nums,
split='train',
normal_channel=False
)
elif args.dataset == 'ShapeNetPart':
DATASET = PartNormalDataset(
root=data_path,
npoint=args.input_point_nums,
split='train',
normal_channel=False
)
else:
raise NotImplementedError
T_DataLoader = torch.utils.data.DataLoader(
DATASET,
batch_size=args.batch_size,
shuffle=False,
num_workers=args.num_workers
)
print('Finish Loading Dataset...')
return T_DataLoader
def data_preprocess(data):
"""Preprocess the given data and label"""
points, target = data
points = points # [B, N, C]
target = target[:, 0] # [B]
points = points.cuda()
target = target.cuda()
return points, target
def save_tensor_as_txt(args, points, filename):
"""Save the torch tensor into a txt file"""
points = points.squeeze(0).detach().cpu().numpy()
file_path = os.path.join(args.output_dir)
if not os.path.exists(file_path):
os.makedirs(file_path)
with open(os.path.join(file_path,filename), "w") as f:
for i in range(points.shape[0]):
msg = str(points[i][0]) + ' ' + str(points[i][1]) + ' ' + str(points[i][2])
f.write(msg+'\n')
f.close()
def rotation_pc(pointcloud, del_1, del_2, del_3):
pointcloud = pointcloud.clone().detach().cpu().numpy()
alpha = np.pi / 180. * del_1
gamma = np.pi / 180. * del_2
beta = np.pi / 180. * del_3
matrix_1 = np.array([[1, 0, 0], [0, np.cos(alpha), -np.sin(alpha)], [0, np.sin(alpha), np.cos(alpha)]])
matrix_2 = np.array([[np.cos(gamma), 0, np.sin(gamma)], [0, 1, 0], [-np.sin(gamma), 0, np.cos(gamma)]])
matrix_3 = np.array([[np.cos(beta), -np.sin(beta), 0], [np.sin(beta), np.cos(beta), 0], [0, 0, 1]])
new_pc = np.matmul(pointcloud, matrix_1)
new_pc = np.matmul(new_pc, matrix_2)
new_pc = np.matmul(new_pc, matrix_3).astype('float32')
return new_pc
def output_delta_list(del_13_list, del_2_list, num_class):
if len(del_13_list) * len(del_2_list) < num_class:
print("Excessive interval! ")
exit(-1)
all_combine_list, del_list = [], []
for i in range(len(del_13_list)):
for j in range(len(del_2_list)):
del_1 = del_13_list[i][0]
del_2 = del_2_list[j]
del_3 = del_13_list[i][1]
all_combine_list.append([del_1, del_2, del_3])
del_list = random.sample(all_combine_list, num_class)
print(del_list)
return del_list
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='PointAPA: Towards Availability Poisoning Attacks in 3D Point Clouds')
parser.add_argument('--batch_size', type=int, default=1, metavar='N', help='input batch size for training (default: 1)')
parser.add_argument('--input_point_nums', type=int, default=1024, help='Point nums of each point cloud')
parser.add_argument('--seed', type=int, default=2022, metavar='S', help='random seed (default: 2022)')
parser.add_argument('--dataset', type=str, default='ModelNet10', choices=['ModelNet10', 'ModelNet40', 'ShapeNetPart'])
parser.add_argument('--num_workers', type=int, default=4, help='Worker nums of data loading.')
parser.add_argument('--normal', action='store_true', default=False, help='Whether to use normal information [default: False]')
parser.add_argument('--interval', type=int, default=42, help='The interval of rotation') #ablation study of interval angle
args = parser.parse_args()
args.device = torch.device("cuda")
set_seed(args.seed)
num_class = 0
if args.dataset == 'ModelNet40':
num_class = 40
data_path = "./clean_data/modelnet40_normal_resampled"
elif args.dataset == 'ShapeNetPart':
num_class = 16
data_path = './clean_data/shapenetcore_partanno_segmentation_benchmark_v0_normal/'
elif args.dataset == 'ModelNet10':
num_class = 10
data_path = "./clean_data/modelnet40_normal_resampled"
assert num_class != 0
args.num_class = num_class
args.output_dir = os.path.join("poison_data", args.dataset, str(args.interval))
del_13_list = [[0,0], [0,10], [0,20], [10,10], [10,20], [20,20]] #alpha and beta angle list
del_2_list = [] #gamma angle list
for i in range(math.ceil(360/args.interval)-1):
del_2_list.append(args.interval*(i+1))
delta_list = output_delta_list(del_13_list, del_2_list, num_class)
dataloader = load_data(args, data_path)
pbar = tqdm(enumerate(dataloader), total=len(dataloader))
start = time.time()
for batch_id, data in pbar:
if args.dataset == 'ShapeNetPart':
data = data[:2]
points, target = data_preprocess(data)
target = target.long()
"""PointAPA poisoning process"""
poi_points = rotation_pc(points, del_1=delta_list[target.item()][0], del_2=delta_list[target.item()][1], del_3=delta_list[target.item()][2])
poi_points = torch.tensor(poi_points).cuda()
#del_2 controls z axis
save_tensor_as_txt(args, poi_points, f'{batch_id}_poi_{target.item()}.txt')
"""visualizing point cloud samples"""
# from utils.visual_util import plot_pcd_three_views
# titles = ['viewpoint 1', 'viewpoint 2', 'viewpoint 3']
# file_path = os.path.join(args.output_dir, 'fig')
# if not os.path.exists(file_path):
# os.makedirs(file_path)
# plot_pcd_three_views(os.path.join(file_path, f'{batch_id}_ori_{target.item()}.png'),[points.squeeze(0).detach().cpu().numpy()],titles)
# plot_pcd_three_views(os.path.join(file_path,f'{batch_id}_poi_{target.item()}.png'),[poi_points.squeeze(0).detach().cpu().numpy()], titles)
end = time.time()
spent_hour, spent_min = transform_time(start, end)
print("The time overhead of PointAPA is {} h {} min".format(spent_hour, spent_min))