-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathTest.py
186 lines (141 loc) · 6.64 KB
/
Test.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
import torch
from torchvision import transforms
from tools.FASDataset import FASDataset
from trainer.FASTTester import FASTTester
import argparse
from utils.utils import read_cfg, get_device, build_network
import os
import numpy as np
from tqdm import tqdm
from tools.FASDataset_RAW import FASDataset_RAW
# from Train import find_mean, training_data_RAW
def testing_data_RAW(cfg):
#train_transform = transforms.Compose([RandomGammaCorrection(max_gamma=cfg['dataset']['augmentation']['gamma_correction'][1],
# min_gamma=cfg['dataset']['augmentation']['gamma_correction'][0]),
# transforms.RandomResizedCrop(cfg['model']['input_size'][0]),
# transforms.RandomRotation(cfg['dataset']['augmentation']['rotation_range']),
# transforms.RandomHorizontalFlip(),
train_transform = transforms.Compose ([transforms.Resize(cfg['model']['input_size']),
transforms.ToTensor(),
transforms.Normalize(cfg['dataset']['mean'], cfg['dataset']['sigma'])])
trainset = FASDataset_RAW(root_dir=cfg['dataset']['root'],
images_file=cfg['dataset']['test_images'],
transform=train_transform,
smoothing=cfg['train']['smoothing'])
trainloader = torch.utils.data.DataLoader(dataset=trainset,
batch_size=cfg['test']['batch_size'],
shuffle=True,
num_workers=2)
return trainloader
def find_mean(trainloader):
# Try to calculate max_d and pm per batch and then calculate the meanof batches
pbar = tqdm(trainloader, total=len(trainloader), dynamic_ncols=True)
x_max = []
x_min = []
y_max = []
y_min = []
z_max = []
z_min = []
all_x = []
all_y = []
all_z = []
for _, point_map, _ in pbar:
# Downsample 10k -> 2.5k ?
max_val = point_map.max(2)[0]
min_val = point_map.min(2)[0]
# print(max_val.shape)
lx = max_val[:, 0]
sx = min_val[:, 0]
ly = max_val[:, 1]
sy = min_val[:, 1]
lz = max_val[:, 2]
sz = min_val[:, 2]
x_max.append(lx)
x_min.append(sx)
y_max.append(ly)
y_min.append(sy)
z_max.append(lz)
z_min.append(sz)
all_x.append(point_map[:, 0])
all_y.append(point_map[:, 1])
all_z.append(point_map[:, 2])
lx = torch.cat(x_max, axis=0).max()
sx = torch.cat(x_min, axis=0).min()
ly = torch.cat(y_max, axis=0).max()
sy = torch.cat(y_min, axis=0).min()
lz = torch.cat(z_max, axis=0).max()
sz = torch.cat(z_min, axis=0).min()
print(len(all_x))
print(len(all_y))
print(len(all_z))
all_x = torch.cat(all_x, axis=0)
all_y = torch.cat(all_y, axis=0)
all_z = torch.cat(all_z, axis=0)
print(f"Check shape: {all_z.shape}")
max_d = max([lx - sx, ly - sy, lz - sz])
pm = torch.tensor([((lx + sx)/2).item(), ((ly + sy)/2).item(),
((lz + sz)/2).item()])
pm = torch.tensor([torch.mean(all_x, dim=(0, 1)), torch.mean(all_y, dim=(0, 1)),
torch.mean(all_z, dim=(0, 1))])
return max_d, np.array(torch.unsqueeze(pm, axis=1))
def test_data(cfg, max_d, pm):
test_transform = transforms.Compose ([transforms.Resize(cfg['model']['input_size']),
transforms.ToTensor(),
transforms.Normalize(cfg['dataset']['mean'], cfg['dataset']['sigma'])])
testset = FASDataset(root_dir=cfg['dataset']['root'],
images_file=cfg['dataset']['test_images'],
transform=test_transform,
smoothing=False,
max_d=max_d, pm=pm,
rot_crop=False)
print(f"Testing on {len(testset)} images")
testloader = torch.utils.data.DataLoader(dataset=testset,
batch_size=cfg['test']['batch_size'],
shuffle=True,
num_workers=cfg['dataset']['num_workers'],
pin_memory=True)
return testloader
def main():
print("Starting testing 3DPC_NET anti-spoofing")
print("Pytorch Version:" + torch.__version__)
print("Cuda Version:" + torch.version.cuda)
#parsing arguments----------------------------------
parser = argparse.ArgumentParser()
parser.add_argument("--load_model", type=str, help="Path where to load model to continue training")
parser.add_argument("--exp_folder", type=str, default="experiments/exp_1", help="Experiment folder to where load files")
parser.add_argument("--csv_path", type=str, default="Protocol1_results.csv", help="Path to where save csv results")
args = parser.parse_args()
#---------------------------------------------------
print("Reading config file....")
cfg_path = os.path.join(args.exp_folder, 'train_config.yaml')
cfg = read_cfg(cfg_file=cfg_path)
print ("ok")
device = get_device(cfg)
print('Using {} device'.format(device))
print("Load Network...")
network = build_network(cfg)
print("ok")
print("model:" + cfg['model']['base'])
# Precisa carregar os valores de normalização utilizados no treino?
#! Normalizar os dados com os valores achados no conjunto de teste?
# print("Finding normalization values in training data...")
# trainloader = testing_data_RAW(cfg)
# max_d, pm = find_mean(trainloader)
# print("Ok")
# print("========= Normalization values =========")
# # TODO: Save normalization values on experiments folder
# print(f"Maximum distance: {max_d}")
# print(f"Medium point: {pm}")
# Remeber point cloud data is still summed by 0.5
# Norm values for OULU Train Protocol 1
# max_d = torch.tensor(1598)
max_d = 1598
pm = np.array([[582.2], [736.4], [300.2]])
# print(pm.shape)
testloader = test_data(cfg, max_d=max_d, pm=pm)
tester = FASTTester(cfg=cfg, network=network, device=device,
testloader=testloader, csv_path=args.csv_path,
exp_folder=args.exp_folder, model_path=args.load_model)
tester.test()
if __name__=='__main__':
main()