SPECTRUM/test.py at main · NiceRingNode/SPECTRUM · GitHub

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
# -*- coding:utf-8 -*-

import torch
import torch.backends.cudnn as cudnn
from torch.utils.data import DataLoader
from model import SPECTRUM,get_len_mask
from dataset import Numbers,TestSampler,collate_fn
import numpy as np
from utils import create_logger,load_ckpt,count_parameters
from dist import dist_skilled_forgery,dist_random_forgery
from verify import verify
import argparse,os,time,pickle,json

torch._C._jit_set_profiling_mode(False)
torch._C._jit_set_profiling_executor(False)
cudnn.enabled = True
cudnn.benchmark = False
cudnn.deterministic = True
torch.cuda.empty_cache()

parser = argparse.ArgumentParser()
parser.add_argument('--train_user_number',type=int,default=202)
parser.add_argument('--seed',type=int,default=123)
parser.add_argument('--cuda',type=bool,default=True)
parser.add_argument('--genuine_sample',type=int,default=20)
parser.add_argument('--forgery_sample',type=int,default=20)
parser.add_argument('--folder',type=str,default='data')
parser.add_argument('--round_num',type=str,default='all')
parser.add_argument('--data_name',type=str,default='real')
parser.add_argument('--weights_root',type=str,default='')
parser.add_argument('--output_root',type=str,default='./output')
parser.add_argument('--log_root',type=str,default='./logs')
parser.add_argument('--gpu',type=int,default=0)
parser.add_argument('--epoch',type=int,default=39)
parser.add_argument('--dropout',type=float,default=0.1)
parser.add_argument('--template_num',type=int,default=4)
parser.add_argument('--d_hidden',type=int,default=128)
parser.add_argument('--test_all_eer',action='store_true')
parser.add_argument('--name',type=str,default='')
parser.add_argument('--notes',type=str,default='')
opt = parser.parse_args()

with open(f'{opt.weights_root}/settings.json','r',encoding='utf-8') as f:
    settings = json.loads(f.read())
opt.seed = settings['seed']
opt.name = settings['name']
opt.notes = settings['notes']
opt.d_hidden = settings['d_hidden']
opt.data_name = settings['data_name']
opt.round_num = settings['round_num']
opt.log_root = settings['log_root']
if opt.round_num == 'all':
    opt.genuine_sample = 20
    opt.forgery_sample = 20
else:
    opt.genuine_sample = 10
    opt.forgery_sample = 10

np.random.seed(opt.seed)
torch.manual_seed(opt.seed)
torch.cuda.manual_seed_all(opt.seed)

os.makedirs(opt.log_root,exist_ok=True)
os.makedirs(opt.output_root,exist_ok=True)

sample_interval = opt.genuine_sample + opt.forgery_sample
logger = create_logger(opt.log_root,name=opt.name,test=True)
with open(f'./{opt.folder}/{opt.data_name}/{opt.round_num}-{opt.data_name}-test-200.pkl','rb') as f:
    handwriting_info = pickle.load(f,encoding='iso-8859-1')
test_dataset = Numbers(handwriting_info,train=False)
users_cnt = test_dataset.config['users_cnt']
test_sampler = TestSampler(users_cnt,opt.genuine_sample,opt.forgery_sample)
test_loader = DataLoader(test_dataset,batch_sampler=test_sampler,collate_fn=collate_fn)

model = SPECTRUM(test_dataset.feature_dims,64,128)


logger.info(f'./{opt.folder}/{opt.data_name}/{opt.round_num}-{opt.data_name}-test-200.pkl\n'
    f'test loader length: {len(test_loader)}\n'
    f'genuine_sample: {opt.genuine_sample}\nforgery_sample: {opt.forgery_sample}\n'
    f'sample_interval: {sample_interval}\nseed: {opt.seed}\n'
    f'model: {model.__class__.__name__}\nnotes: {opt.notes}')

if opt.cuda and torch.cuda.is_available():
    torch.cuda.set_device(int(opt.gpu))
    device = torch.device(f'cuda:{opt.gpu}')
else:
    device = torch.device('cpu')
model = model.to(device)

@torch.no_grad()
def test():
    logger.info('<======test begins.======>')
    model.eval()
    output,output_freq = [],[]
    start = time.time()
    for i,(features,features_lens,_,_) in enumerate(test_loader):
        features = torch.from_numpy(features).to(device)
        features_lens = torch.tensor(features_lens).long().to(device)
        mask = get_len_mask(features_lens).to(device)
        y_vector,_,_,y_freq = model(features,mask)
        output.append(y_vector.cpu().numpy())
        output_freq.append(y_freq.cpu().numpy())
    end = time.time()
    extraction_period = end - start
    logger.info(f'feature extraction time per sample: {1000 * extraction_period / len(test_dataset):.3f} ms/s')

    skilled_start = time.time()
    dist_genuine,dist_forgery,dist_template,dfg,dff,dft = dist_skilled_forgery(output,output_freq,opt.template_num,0,opt.genuine_sample,opt.forgery_sample)
    np.save(f'{opt.output_root}/dist_genuine.npy',dist_genuine)
    np.save(f'{opt.output_root}/dist_forgery.npy',dist_forgery)
    np.save(f'{opt.output_root}/dist_template.npy',dist_template)
    np.save(f'{opt.output_root}/dist_fgenuine.npy',dfg)
    np.save(f'{opt.output_root}/dist_fforgery.npy',dff)
    np.save(f'{opt.output_root}/dist_ftemplate.npy',dft)
    verify(logger,opt.template_num,users_cnt,opt.genuine_sample,opt.forgery_sample,rf=False)
    skilled_end = time.time()
    skilled_period = skilled_end - skilled_start
    logger.info(f'skilled forgery verification time per sample: {1000 * skilled_period / len(test_dataset):.3f} ms/s')
    logger.info(f'skilled forgery verification time per sample (total): {1000 * (skilled_period + extraction_period) / len(test_dataset):.3f} ms/s')

    random_start = time.time()
    dist_genuine,dist_forgery,dist_template,dfg,dff,dft = dist_random_forgery(output,output_freq,opt.template_num,0,opt.genuine_sample)
    np.save(f'{opt.output_root}/dist_genuine.npy',dist_genuine)
    np.save(f'{opt.output_root}/dist_forgery.npy',dist_forgery)
    np.save(f'{opt.output_root}/dist_template.npy',dist_template)
    np.save(f'{opt.output_root}/dist_fgenuine.npy',dfg)
    np.save(f'{opt.output_root}/dist_fforgery.npy',dff)
    np.save(f'{opt.output_root}/dist_ftemplate.npy',dft)
    verify(logger,opt.template_num,users_cnt,opt.genuine_sample,opt.forgery_sample,rf=True)
    random_end = time.time()
    random_period = random_end - random_start
    logger.info(f'random forgery verification time per sample: {1000 * random_period / len(test_dataset):.3f} ms/s')
    logger.info(f'random forgery verification time per sample (total): {1000 * (random_period + extraction_period) / len(test_dataset):.3f} ms/s')

    _,params = count_parameters(model)
    logger.info(f'params: {params / 1e6:.6f}')

@torch.no_grad()
def test_all(template_num):
    logger.info('<======test all begins.======>')
    model.eval()
    output,output_freq = [],[]
    for i,(features,features_lens,_,_) in enumerate(test_loader):
        features = torch.from_numpy(features).to(device)
        features_lens = torch.tensor(features_lens).long().to(device)
        mask = get_len_mask(features_lens).to(device)
        y_vector,_,_,y_freq = model(features,mask)
        output.append(y_vector.cpu().numpy())
        output_freq.append(y_freq.cpu().numpy())

    if template_num == 4 or template_num == 1:
        dist_genuine,dist_forgery,dist_template,dfg,dff,dft = dist_skilled_forgery(output,output_freq,template_num,
            0,opt.genuine_sample,opt.forgery_sample)
        np.save(f'{opt.output_root}/dist_genuine.npy',dist_genuine)
        np.save(f'{opt.output_root}/dist_forgery.npy',dist_forgery)
        np.save(f'{opt.output_root}/dist_template.npy',dist_template)
        np.save(f'{opt.output_root}/dist_fgenuine.npy',dfg)
        np.save(f'{opt.output_root}/dist_fforgery.npy',dff)
        np.save(f'{opt.output_root}/dist_ftemplate.npy',dft)
        verify(logger,template_num,users_cnt,opt.genuine_sample,opt.forgery_sample,rf=False)

        dist_genuine,dist_forgery,dist_template,dfg,dff,dft = dist_random_forgery(output,output_freq,template_num,
            0,opt.genuine_sample)
        np.save(f'{opt.output_root}/dist_genuine.npy',dist_genuine)
        np.save(f'{opt.output_root}/dist_forgery.npy',dist_forgery)
        np.save(f'{opt.output_root}/dist_template.npy',dist_template)
        np.save(f'{opt.output_root}/dist_fgenuine.npy',dfg)
        np.save(f'{opt.output_root}/dist_fforgery.npy',dff)
        np.save(f'{opt.output_root}/dist_ftemplate.npy',dft)
        verify(logger,template_num,users_cnt,opt.genuine_sample,opt.forgery_sample,rf=True)
    else:
        dist_genuine,dist_forgery,dist_template,dfg,dff,dft = dist_skilled_forgery(output,output_freq,template_num,
            0,opt.genuine_sample,opt.forgery_sample)
        np.save(f'{opt.output_root}/dist_genuine.npy',dist_genuine)
        np.save(f'{opt.output_root}/dist_forgery.npy',dist_forgery)
        np.save(f'{opt.output_root}/dist_template.npy',dist_template)
        np.save(f'{opt.output_root}/dist_fgenuine.npy',dfg)
        np.save(f'{opt.output_root}/dist_fforgery.npy',dff)
        np.save(f'{opt.output_root}/dist_ftemplate.npy',dft)
        verify(logger,template_num,users_cnt,opt.genuine_sample,opt.forgery_sample,rf=False)

def main():
    if not opt.test_all_eer:
        load_ckpt(model,f'{opt.weights_root}/ckpt-{opt.epoch}-{opt.name}.pth',device,logger,mode='test')
        time_elapsed_start = time.time()
        test()
        logger.info(f'time elapsed: {time.time() - time_elapsed_start:.5f}s\n')
    else:
        load_ckpt(model,f'{opt.weights_root}/ckpt-{opt.epoch}-{opt.name}.pth',device,logger,mode='test')
        for i in range(4,0,-1):
            time_elapsed_start = time.time()
            test_all(i)
            logger.info(f'time elapsed: {time.time() - time_elapsed_start:.5f}s\n')

if __name__ == '__main__':
    main()