/
trainer_optimizer.py
213 lines (182 loc) · 10.2 KB
/
trainer_optimizer.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
import os
import time
from collections import OrderedDict
import torch
import torch.distributed as dist
import torch.optim
import torchvision.utils as vutils
from torch.utils.data import DataLoader
from data.CRN_dataset import CRNShapeNet
from data.ply_dataset import PlyDataset, RealDataset, GeneratedDataset
from arguments import Arguments
from loss import *
from optde import OptDE
from model.network import Generator, Discriminator
from external.ChamferDistancePytorch.chamfer_python import distChamfer, distChamfer_raw
import random
import numpy as np
def seed_torch(seed=1029):
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True
seed_torch()
time_stamp = time.strftime('Log_%Y-%m-%d_%H-%M-%S/', time.gmtime())
class Trainer(object):
def __init__(self, args):
self.args = args
save_inversion_dirname = args.save_inversion_path.split('/')
log_pathname = './'+args.log_dir+'/'+ save_inversion_dirname[-3] + '/' + save_inversion_dirname[-2] + '/log.txt'
args.log_pathname = log_pathname
self.model = OptDE(self.args)
###Load Virtual Train Data
self.virtual_data_name = self.args.virtualdataset
self.args.dataset = self.virtual_data_name
if self.virtual_data_name in ['ScanNet', 'MatterPort']:
self.args.split = 'trainval'
elif self.virtual_data_name in ['ModelNet', '3D_FUTURE', 'KITTI', 'CRN']:
self.args.split = 'train'
if self.virtual_data_name in ['MatterPort','ScanNet','KITTI','PartNet']:
train_dataset = PlyDataset(self.args)
elif self.virtual_data_name in ['ModelNet', '3D_FUTURE']:
train_dataset = GeneratedDataset(self.args)
else:
train_dataset = CRNShapeNet(self.args)
p2c_batch_size = 10#20
self.train_dataloader = DataLoader(
train_dataset,
batch_size=p2c_batch_size,
shuffle=False,
pin_memory=True)
###Load Virtual Test Data
self.args.split = 'test'
if self.virtual_data_name in ['MatterPort','ScanNet','KITTI','PartNet']:
test_dataset = PlyDataset(self.args)
elif self.virtual_data_name in ['ModelNet', '3D_FUTURE']:
test_dataset = GeneratedDataset(self.args)
else:
test_dataset = CRNShapeNet(self.args)
self.test_dataloader = DataLoader(
test_dataset,
batch_size=1,
shuffle=False,
pin_memory=True)
###Load Real Train Data
self.real_data_name = self.args.realdataset
self.args.dataset = self.real_data_name
if self.real_data_name in ['ScanNet', 'MatterPort']:
self.args.split = 'trainval'
elif self.real_data_name in ['ModelNet', '3D_FUTURE', 'KITTI', 'CRN']:
self.args.split = 'train'
if self.real_data_name in ['MatterPort','ScanNet','KITTI','PartNet']:
real_train_dataset = RealDataset(self.args)
elif self.real_data_name in ['ModelNet', '3D_FUTURE']:
real_train_dataset = GeneratedDataset(self.args)
elif self.real_data_name in ['CRN']:
real_train_dataset = CRNShapeNet(self.args)
self.real_train_dataloader = DataLoader(
real_train_dataset,
batch_size=p2c_batch_size,
shuffle=False,
pin_memory=True)
###Load Real Test Data
self.args.split = 'test'
if self.real_data_name in ['MatterPort','ScanNet','KITTI','PartNet']:
real_test_dataset = RealDataset(self.args)
elif self.real_data_name in ['ModelNet', '3D_FUTURE']:
real_test_dataset = GeneratedDataset(self.args)
elif self.real_data_name in ['CRN']:
real_test_dataset = CRNShapeNet(self.args)
self.real_test_dataloader = DataLoader(
real_test_dataset,
batch_size=1,
shuffle=False,
pin_memory=True)
def train(self):
load_path_name = self.args.finetune_ckpt_load
print(load_path_name)
test_real_ucd_loss_list = []
test_real_uhd_loss_list = []
test_real_cd_loss_list = []
for i, data in enumerate(self.real_test_dataloader):
tic = time.time()
# with gt
if self.real_data_name in ['ScanNet', 'MatterPort', 'KITTI']:
partial, index = data
elif self.real_data_name in ['ModelNet', '3D_FUTURE', 'CRN']:
gt, partial, index = data
gt = gt.squeeze(0).float().cuda()
partial = partial.squeeze(0).float().cuda()
# reset G for each new input
#self.model.reset_G(pcd_id=index.item())
self.model.reset_G_tmp()
self.model.pcd_id = index[0].item()
# set target and complete shape
# for ['reconstruction', 'jittering', 'morphing'], GT is used for reconstruction
# else, GT is not involved for training
if self.real_data_name in ['ScanNet', 'MatterPort', 'KITTI']:
self.model.set_target(partial=partial)
elif self.real_data_name in ['ModelNet', '3D_FUTURE', 'CRN']:
self.model.set_target(gt=gt, partial=partial)
# inversion
self.model.reset_whole_network(load_path_name)
if self.real_data_name in ['ScanNet', 'MatterPort', 'KITTI']:
test_real_ucd_loss, test_real_uhd_loss = self.model.finetune()
elif self.real_data_name in ['ModelNet', '3D_FUTURE', 'CRN']:
test_real_ucd_loss, test_real_uhd_loss, test_real_cd_loss = self.model.finetune(bool_gt=True)
test_real_cd_loss_list.append(test_real_cd_loss)
test_real_ucd_loss_list.append(test_real_ucd_loss)
test_real_uhd_loss_list.append(test_real_uhd_loss)
toc = time.time()
print(i, 'out of', len(self.real_test_dataloader),'done in ',int(toc-tic),'s')
if self.real_data_name in ['ModelNet', '3D_FUTURE', 'CRN']:
np.save(self.args.save_inversion_path+'/cd_list.npy', np.array(test_real_cd_loss_list))
test_real_cd_loss_mean = np.mean(np.array(test_real_cd_loss_list))
test_real_ucd_loss_mean = np.mean(np.array(test_real_ucd_loss_list))
test_real_uhd_loss_mean = np.mean(np.array(test_real_uhd_loss_list))
if self.real_data_name in ['ModelNet', '3D_FUTURE', 'CRN']:
print("Mean CD on Real Test Set:", test_real_cd_loss_mean)
print("Mean UCD on Real Test Set:", test_real_ucd_loss_mean)
print("Mean UHD on Real Test Set:", test_real_uhd_loss_mean)
with open(self.args.log_pathname, "a") as file_object:
if self.real_data_name in ['ModelNet', '3D_FUTURE', 'CRN']:
msg = "Mean CD on Real Test Set:" + "%.8f"%test_real_cd_loss_mean
file_object.write(msg+'\n')
msg = "Mean UCD on Real Test Set:" + "%.8f"%test_real_ucd_loss_mean
file_object.write(msg+'\n')
msg = "Mean UHD on Real Test Set:" + "%.8f"%test_real_uhd_loss_mean
file_object.write(msg+'\n')
os.system("mv " + './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-3] + '/' + time_stamp[:-1] + '/saved_results/*' + ' ./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-3] + '/' + time_stamp[:-1] + '/best_results/')
if __name__ == "__main__":
args = Arguments(stage='inversion').parser().parse_args()
args.device = torch.device('cuda:'+str(args.gpu) if torch.cuda.is_available() else 'cpu')
torch.cuda.set_device(args.device)
if not os.path.isdir('./'+args.log_dir+'/'):
os.mkdir('./'+args.log_dir+'/')
if not os.path.isdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1]):
os.mkdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1])
if not os.path.isdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1]):
os.mkdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1])
if not os.path.isdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/best_results'):
os.mkdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1]+'/best_results')
if not os.path.isdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code'):
os.mkdir('./'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code')
os.system('cp %s %s'% ('run_optimizer.sh', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] +
'/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('trainer.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('trainer_optimizer.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-
1] + '/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('optde.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('model/network.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('data/ply_dataset.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('data/real_dataset.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code/'))
os.system('cp %s %s'% ('trainer_optimizer.py', './'+args.log_dir+'/' + args.save_inversion_path.split('/')[-1] + '/' + time_stamp[:-1] + '/code/'))
args.save_inversion_path += '/' + time_stamp[:-1]
args.ckpt_path_name = args.save_inversion_path + '/' + args.class_choice + '.pt'
args.save_inversion_path += '/' + 'saved_results'
trainer = Trainer(args)
trainer.train()