-
Notifications
You must be signed in to change notification settings - Fork 3
/
config.py
214 lines (172 loc) · 7.07 KB
/
config.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
import argparse
arg_lists = []
parser = argparse.ArgumentParser()
def add_argument_group(name):
arg = parser.add_argument_group(name)
arg_lists.append(arg)
return arg
def str2bool(v):
return v.lower() in ("true", "1")
# -----------------------------------------------------------------------------
# Data
data_arg = add_argument_group("Data")
data_arg.add_argument(
"--data_pre", type=str, default="data", help=""
"prefix for the dump folder locations")
data_arg.add_argument(
"--data_tr", type=str, default="sun3d", help=""
"name of the dataset for train")
data_arg.add_argument(
"--data_va", type=str, default="sun3d", help=""
"name of the dataset for valid")
data_arg.add_argument(
"--data_te", type=str, default="sun3d", help=""
"name of the dataset for test")
# -----------------------------------------------------------------------------
# Network
net_arg = add_argument_group("Network")
net_arg.add_argument(
"--net_depth", type=int, default=12, help=""
"number of layers")
net_arg.add_argument(
"--net_nchannel", type=int, default=128, help=""
"number of channels in a layer")
net_arg.add_argument(
"--net_act_pos", type=str, default="post",
choices=["pre", "mid", "post"], help=""
"where the activation should be in case of resnet")
net_arg.add_argument(
"--net_gcnorm", type=str2bool, default=True, help=""
"whether to use context normalization for each layer")
net_arg.add_argument(
"--net_batchnorm", type=str2bool, default=True, help=""
"whether to use batch normalization")
net_arg.add_argument(
"--net_bn_test_is_training", type=str2bool, default=False, help=""
"is_training value for testing")
net_arg.add_argument(
"--net_concat_post", type=str2bool, default=False, help=""
"retrieve top k values or concat from different layers")
net_arg.add_argument(
"--gpu_options", type=str, default='gpu', choices=['gpu', 'cpu'],
help="choose which gpu or cpu")
net_arg.add_argument(
"--gpu_number", type=str, default='0',
help="choose which gpu number")
net_arg.add_argument(
"--post_refine", type=str2bool, default=False, help="use the post refinement or not")
net_arg.add_argument(
"--R_method", type=str, default='svd', choices=['svd', 'regression'],
help="choose which estimation method for rotation")
net_arg.add_argument(
"--PCFD_layer", type=int, default=10,
help="the layer number of PCFD module")
net_arg.add_argument(
"--SCA_layer", type=int, default=4,
help="the layer number of SCA module")
# -----------------------------------------------------------------------------
# Loss
loss_arg = add_argument_group("loss")
loss_arg.add_argument(
"--loss_decay", type=float, default=0.0, help=""
"l2 decay")
loss_arg.add_argument(
"--loss_classif", type=float, default=0.5, help=""
"weight of the classification loss")
loss_arg.add_argument(
"--loss_reconstruction", type=float, default=0.01, help=""
"weight of the essential loss")
loss_arg.add_argument(
"--loss_reconstruction_init_iter", type=int, default=0, help=""
"initial iterations to run only the classification loss")
loss_arg.add_argument(
"--drift_loss_type", type=str, default='l1', choices=['alignment', 'l1'],
help = "choose which form of loss function to penalizes the middle-layer translation")
# -----------------------------------------------------------------------------
# Training
train_arg = add_argument_group("Train")
train_arg.add_argument(
"--run_mode", type=str, default="train", help=""
"run_mode")
train_arg.add_argument(
"--train_batch_size", type=int, default=16, help=""
"batch size")
train_arg.add_argument(
"--train_max_tr_sample", type=int, default=10000, help=""
"number of max training samples")
train_arg.add_argument(
"--train_max_va_sample", type=int, default=1000, help=""
"number of max validation samples")
train_arg.add_argument(
"--train_max_te_sample", type=int, default=1000, help=""
"number of max test samples")
train_arg.add_argument(
"--train_lr", type=float, default=1e-5, help=""
"learning rate")
train_arg.add_argument(
"--train_epoch", type=int, default=200, help=""
"training iterations to perform")
train_arg.add_argument(
"--train_step", type=int, default=200, help=""
"training iterations to perform")
train_arg.add_argument(
"--res_dir", type=str, default="./logs", help=""
"base directory for results")
train_arg.add_argument(
"--log_dir", type=str, default="logs_lie", help=""
"save directory name inside results")
train_arg.add_argument(
"--test_log_dir", type=str, default="", help=""
"which directory to test inside results")
train_arg.add_argument(
"--val_intv", type=int, default=5, help=""
"validation interval")
train_arg.add_argument(
"--report_intv", type=int, default=100, help=""
"summary interval")
net_arg.add_argument(
"--loss_function", type=str, default='l1', choices=['l1', 'l2', 'wls', 'gm', 'l05'],
help="choose which loss function")
test_arg = add_argument_group("Test")
test_arg.add_argument(
"--R_thre", type=float, default=5, help="the rotation threshold for computing the registration recall and map")
test_arg.add_argument(
"--t_thre", type=float, default=15, help="the translation threshold for computing the registration recall and map")
# -----------------------------------------------------------------------------
# Data Augmentation
d_aug = add_argument_group('Augmentation')
d_aug.add_argument("--data_aug", type=str2bool, default=False, help="Perform data Augmentation")
d_aug.add_argument("--aug_cl", type=str2bool, default=True, help="Perform Curriculum Learning")
d_aug.add_argument("--aug_dir", type=str, default="augmentented", help="save directory name inside results")
# -----------------------------------------------------------------------------
# Visualization
vis_arg = add_argument_group('Visualization')
vis_arg.add_argument(
"--tqdm_width", type=int, default=79, help=""
"width of the tqdm bar")
vis_arg.add_argument(
"--reg_flag", type=str2bool, default=False, help="Refine transformation")
test_arg = add_argument_group('Test')
test_arg.add_argument(
"--reg_function", type=str, default='fast', choices=['fast', 'global'], help="Registration function: global or fast")
test_arg.add_argument(
"--representation", type=str, default='lie', choices=['lie', 'quat', 'linear'], help="Type of Representation")
def setup_dataset(dataset_name):
dataset_name = dataset_name.split(".")
data_dir = []
for name in dataset_name:
data_dir.append(name)
assert data_dir
return data_dir
def get_config():
config, unparsed = parser.parse_known_args()
# Setup the dataset related things
for _mode in ["tr", "va", "te"]:
data_dir = setup_dataset(
getattr(config, "data_" + _mode))
setattr(config, "data_dir_" + _mode, data_dir)
# setattr(config, "data_geom_type_" + _mode, geom_type)
# setattr(config, "data_vis_th_" + _mode, vis_th)
return config, unparsed
def print_usage():
parser.print_usage()