-
Notifications
You must be signed in to change notification settings - Fork 39
/
train.py
executable file
·397 lines (326 loc) · 16.6 KB
/
train.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
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
import argparse
import os
import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.optim as optim
from torch.utils.data import DataLoader
from torch.autograd import Variable
import torch.nn.functional as F
import numpy as np
import time
from tensorboardX import SummaryWriter
from datasets import find_dataset_def
from models import *
from utils import *
import gc
import sys
import datetime
from torchvision import models
from torchvision import utils as vutils
from torchvision import transforms
cudnn.benchmark = True
parser = argparse.ArgumentParser(description='A PyTorch Implementation of MVSNet')
parser.add_argument('--mode', default='train', help='train or test', choices=['train', 'test', 'profile'])
parser.add_argument('--model', default='mvsnet', help='select model')
parser.add_argument('--dataset', default='dtu_yao', help='select dataset')
parser.add_argument('--trainpath', help='train datapath')
parser.add_argument('--testpath', help='test datapath')
parser.add_argument('--trainlist', help='train list')
parser.add_argument('--testlist', help='test list')
parser.add_argument('--epochs', type=int, default=10, help='number of epochs to train')
parser.add_argument('--lr', type=float, default=0.001, help='learning rate')
parser.add_argument('--lrepochs', type=str, default="1,3,5,7:2", help='epoch ids to downscale lr and the downscale rate')
parser.add_argument('--wd', type=float, default=0.0, help='weight decay')
parser.add_argument('--batch_size', type=int, default=12, help='train batch size')
parser.add_argument('--numdepth', type=int, default=192, help='the number of depth values')
parser.add_argument('--interval_scale', type=float, default=1.06, help='the number of depth values')
parser.add_argument('--loadckpt', default=None, help='load a specific checkpoint')
parser.add_argument('--logdir', default='./checkpoints/debug', help='the directory to save checkpoints/logs')
parser.add_argument('--resume',action='store_true', help='continue to train the model') #default=True
parser.add_argument('--summary_freq', type=int, default=20, help='print and summary frequency')
parser.add_argument('--save_freq', type=int, default=1, help='save checkpoint frequency')
parser.add_argument('--seed', type=int, default=1, metavar='S', help='random seed')
# parse arguments and check
args = parser.parse_args()
if args.resume:
assert args.mode == "train"
assert args.loadckpt is None
if args.testpath is None:
args.testpath = args.trainpath
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
# create logger for mode "train" and "testall"
if args.mode == "train":
if not os.path.isdir(args.logdir):
os.mkdir(args.logdir)
current_time_str = str(datetime.datetime.now().strftime('%Y%m%d_%H%M%S'))
print("current time", current_time_str)
print("creating new summary file")
logger = SummaryWriter(args.logdir)
print("argv:", sys.argv[1:])
print_args(args)
# dataset, dataloader
MVSDataset = find_dataset_def(args.dataset)
train_dataset = MVSDataset(args.trainpath, args.trainlist, "train", 3, args.numdepth, args.interval_scale)
test_dataset = MVSDataset(args.testpath, args.testlist, "test", 5, args.numdepth, args.interval_scale)
TrainImgLoader = DataLoader(train_dataset, args.batch_size, shuffle=True, num_workers=8, drop_last=True)
TestImgLoader = DataLoader(test_dataset, args.batch_size, shuffle=False, num_workers=4, drop_last=False)
# class vggNet(nn.Module):
# def __init__(self, pretrained=True):
# super(vggNet, self).__init__()
# self.net = models.vgg16(pretrained=True).features.eval()
# def forward(self, x):
# out = []
# for i in range(len(self.net)):
# x = self.net[i](x)
# #if i in [3, 8, 15, 22, 29]:
# if i in [3, 8, 15]: #提取1，1/2，1/4的特征图
# # print(self.net[i])
# out.append(x)
# return out
# model, optimizer
model = MVSNet(refine=False)
model_feature = vggNet()
if args.mode in ["train", "test"]:
model = nn.DataParallel(model)
model_feature = nn.DataParallel(model_feature)
model.cuda()
model_feature.cuda()
model_loss = mvsnet_loss
optimizer = optim.Adam(model.parameters(), lr=args.lr, betas=(0.9, 0.999), weight_decay=args.wd)
# load parameters
start_epoch = 0
if (args.mode == "train" and args.resume) or (args.mode == "test" and not args.loadckpt):
saved_models = [fn for fn in os.listdir(args.logdir) if fn.endswith(".ckpt")]
saved_models = sorted(saved_models, key=lambda x: int(x.split('_')[-1].split('.')[0]))
# use the latest checkpoint file
loadckpt = os.path.join(args.logdir, saved_models[-1])
print("resuming", loadckpt)
state_dict = torch.load(loadckpt)
model.load_state_dict(state_dict['model'])
optimizer.load_state_dict(state_dict['optimizer'])
start_epoch = state_dict['epoch'] + 1
elif args.loadckpt:
# load checkpoint file specified by args.loadckpt
print("loading model {}".format(args.loadckpt))
state_dict = torch.load(args.loadckpt)
model.load_state_dict(state_dict['model'])
print("start at epoch {}".format(start_epoch))
print('Number of model parameters: {}'.format(sum([p.data.nelement() for p in model.parameters()])))
# main function
def train():
milestones = [int(epoch_idx) for epoch_idx in args.lrepochs.split(':')[0].split(',')]
lr_gamma = 1 / float(args.lrepochs.split(':')[1])
lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones, gamma=lr_gamma,
last_epoch=start_epoch - 1)
for epoch_idx in range(start_epoch, args.epochs):
print('Epoch {}:'.format(epoch_idx))
lr_scheduler.step()
global_step = len(TrainImgLoader) * epoch_idx
#training
for batch_idx, sample in enumerate(TrainImgLoader):
start_time = time.time()
global_step = len(TrainImgLoader) * epoch_idx + batch_idx
do_summary = global_step % args.summary_freq == 0
#loss, scalar_outputs, image_outputs = train_sample(sample, detailed_summary=do_summary)
loss,scalar_outputs,image_outputs = train_sample(sample, detailed_summary=do_summary)
if do_summary:
print("save_scalar")
save_scalars(logger, 'train', scalar_outputs, global_step)
print("save_images")
save_images(logger, 'train', image_outputs, global_step)
del scalar_outputs, image_outputs
print(
'Epoch {}/{}, Iter {}/{}, train loss = {:.3f}, time = {:.3f}, GS={}'.format(epoch_idx, args.epochs, batch_idx,
len(TrainImgLoader), loss,
time.time() - start_time,global_step))
# checkpoint
if (epoch_idx + 1) % args.save_freq == 0:
torch.save({
'epoch': epoch_idx,
'model': model.state_dict(),
'optimizer': optimizer.state_dict()},
"{}/model_{:0>6}.ckpt".format(args.logdir, epoch_idx))
# testing
avg_test_scalars = DictAverageMeter()
for batch_idx, sample in enumerate(TestImgLoader):
start_time = time.time()
global_step = len(TrainImgLoader) * epoch_idx + batch_idx
do_summary = global_step % args.summary_freq == 0
loss, scalar_outputs, image_outputs = test_sample(sample, detailed_summary=do_summary)
if do_summary:
save_scalars(logger, 'test', scalar_outputs, global_step)
save_images(logger, 'test', image_outputs, global_step)
avg_test_scalars.update(scalar_outputs)
del scalar_outputs, image_outputs
print('Epoch {}/{}, Iter {}/{}, test loss = {:.3f}, time = {:3f}'.format(epoch_idx, args.epochs, batch_idx,
len(TestImgLoader), loss,
time.time() - start_time))
save_scalars(logger, 'fulltest', avg_test_scalars.mean(), global_step)
print("avg_test_scalars:", avg_test_scalars.mean())
# gc.collect()
def test():
avg_test_scalars = DictAverageMeter()
for batch_idx, sample in enumerate(TestImgLoader):
start_time = time.time()
loss, scalar_outputs, image_outputs = test_sample(sample, detailed_summary=True)
avg_test_scalars.update(scalar_outputs)
del scalar_outputs, image_outputs
print('Iter {}/{}, test loss = {:.3f}, time = {:3f}'.format(batch_idx, len(TestImgLoader), loss,
time.time() - start_time))
if batch_idx % 100 == 0:
print("Iter {}/{}, test results = {}".format(batch_idx, len(TestImgLoader), avg_test_scalars.mean()))
print("final", avg_test_scalars)
def train_sample(sample, detailed_summary=False):
model.train()
optimizer.zero_grad()
sample_cuda = tocuda(sample)
depth_gt = sample_cuda["depth"]
#print("depth_gt_shape:{}".format(depth_gt.shape)) #Batchsize*128*160
mask_gt = sample_cuda["mask"]
intrinsics=sample_cuda["intrinsics"]
extrinsics=sample_cuda["extrinsics"]
#print(sample_cuda["imgs"].shape)
print("Begin forward")
outputs = model(sample_cuda["imgs"], sample_cuda["proj_matrices"], sample_cuda["depth_values"])
depth_est = outputs["depth"]
photometric_confidence = outputs["photometric_confidence"]
#提取特征
# with torch.no_grid()
with torch.no_grad():
print("Begin VGG16 extract feature")
#print(sample_cuda["imgs"].shape) #4*3*3*512*640
imgs_features = sample_cuda["imgs"].clone()
imgs_features=torch.unbind(imgs_features,1) # nivews个4*3*512*640
mean=torch.tensor([[[0.485]],[[0.456]],[[0.406]]],device='cuda')
std=torch.tensor([[[0.229]],[[0.224]],[[0.225]]],device='cuda')
imgs_features = [(img-mean)/std for img in imgs_features]
#normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])
#print(len(imgs_features)) 3
#print(imgs_features[0].shape) 4*3*512*640
outputs_feature = [model_feature(img) for img in imgs_features]
#print(len(outputs_feature)) #3
#print(len(outputs_feature[0])) #1
#print(outputs_feature[0][0].shape) #4*256*128*160
#print(len(outputs_feature)) 3
#print(outputs_feature[0][0].shape) #4*64*512*640-----3
#print(outputs_feature[0][1].shape) #4*28*258*320-----8
#print(outputs_feature[0][2].shape) #4*256*128*160----15
#depth_est=outputs["refined_depth"]
#depth_est=depth_est.squeeze(1)
mask_photometric=photometric_confidence>0.5 #unit8
mask_final = mask_gt*mask_photometric.float()
#print(mask_final)
#print(mask_final.shape)
#print("depth_est_shape:{}".format(depth_est.shape)) #4*128*160
#print(depth_est.shape) #B*128*160
#print(intrinsics.shape) #B*3*3
#print(sample_cuda["imgs"].shape) #B*3*3*512*640
#loss = model_loss(depth_est, depth_gt, mask)
print("Begin calculate loss")
loss,loss_s,loss_photo,loss_ssim,mask_calculate,mask_num,loss_perceptual,loss_normal,normal_by_depth,error_depth_by_normal,depth_by_normal=model_loss(depth_est,intrinsics,extrinsics,sample_cuda["imgs"],mask_photometric,outputs_feature)
print("Begin loss backward")
loss.backward()
print("Begin optimizer")
optimizer.step()
# print(depth_est.shape)
# print(type(depth_est))
# print(depth_by_normal.shape)
# print(type(depth_by_normal))
#print(mask_num)
scalar_outputs = {"loss": loss,"loss_s":loss_s,"loss_photo":loss_photo,"loss_ssim":loss_ssim,"loss_perceptual":loss_perceptual,"loss_normal":loss_normal}
image_outputs = {"depth_est_gt": depth_est * mask_gt, "depth_est_final": depth_est * mask_final,
"depth_est":depth_est,
"depth_by_normal_final":depth_by_normal * mask_final,
"depth_by_normal":depth_by_normal,
"depth_gt": sample["depth"],
"ref_img": sample["imgs"][:, 0],
"mask_gt": sample["mask"],
"photometric":photometric_confidence,
"mask_photometric":mask_photometric.float(),
"normal_by_depth":normal_by_depth.permute(0,3,1,2)
}
#"mask_calculate":mask_calculate}
if detailed_summary:
image_outputs["errormap_gt"] = (depth_est - depth_gt).abs() * mask_gt
image_outputs["errormap_final"] = (depth_est - depth_gt).abs() * mask_final
image_outputs["errormap_depth_normal_final"]=error_depth_by_normal * mask_final
#print(depth_est.device)
#print(depth_gt.device)
#print(mask_final.device)
scalar_outputs["abs_depth_error"] = AbsDepthError_metrics(depth_est, depth_gt, mask_final > 0.5)
scalar_outputs["thres2mm_error"] = Thres_metrics(depth_est, depth_gt, mask_final > 0.5, 2)
scalar_outputs["thres4mm_error"] = Thres_metrics(depth_est, depth_gt, mask_final > 0.5, 4)
scalar_outputs["thres8mm_error"] = Thres_metrics(depth_est, depth_gt, mask_final > 0.5, 8)
#print("loss:{}".format(loss))
#print("abs:{}".format(AbsDepthError_metrics(depth_est, depth_gt, mask > 0.5)))
#print("2:{}".format(Thres_metrics(depth_est, depth_gt, mask > 0.5, 2)))
#print("4:{}".format(Thres_metrics(depth_est, depth_gt, mask > 0.5, 4)))
#print("8:{}".format(Thres_metrics(depth_est, depth_gt, mask > 0.5, 8)))
return tensor2float(loss), tensor2float(scalar_outputs), image_outputs
def test_loss_calculate(depth_est, depth_gt, mask_final):
mask_final=mask_final>0.5
return F.smooth_l1_loss(depth_est[mask_final], depth_gt[mask_final], size_average=True)
@make_nograd_func
def test_sample(sample, detailed_summary=True):
model.eval()
sample_cuda = tocuda(sample)
depth_gt = sample_cuda["depth"]
mask_gt = sample_cuda["mask"]
outputs = model(sample_cuda["imgs"], sample_cuda["proj_matrices"], sample_cuda["depth_values"])
depth_est = outputs["depth"]
photometric_confidence = outputs["photometric_confidence"]
#depth_est=outputs["refined_depth"]
#depth_est=depth_est.squeeze(1)
mask_photometric=photometric_confidence>0.5
mask_final=mask_gt*mask_photometric.float()
loss = test_loss_calculate(depth_est, depth_gt, mask_final)
scalar_outputs = {"loss": loss}
image_outputs = {"depth_est_gt": depth_est * mask_gt, "depth_est_final": depth_est * mask_final,
"depth_gt": sample["depth"],
"ref_img": sample["imgs"][:, 0],
"mask_gt": sample["mask"],
"photometric":photometric_confidence,
"mask_photometric":mask_photometric.float()}
if detailed_summary:
image_outputs["errormap_gt"] = (depth_est - depth_gt).abs() * mask_gt
image_outputs["errormap_final"] = (depth_est - depth_gt).abs() * mask_final
scalar_outputs["abs_depth_error"] = AbsDepthError_metrics(depth_est, depth_gt, mask_final > 0.5)
scalar_outputs["thres2mm_error"] = Thres_metrics(depth_est, depth_gt, mask_final > 0.5, 2)
scalar_outputs["thres4mm_error"] = Thres_metrics(depth_est, depth_gt, mask_final > 0.5, 4)
scalar_outputs["thres8mm_error"] = Thres_metrics(depth_est, depth_gt, mask_final > 0.5, 8)
return tensor2float(loss), tensor2float(scalar_outputs), image_outputs
def profile():
warmup_iter = 5
iter_dataloader = iter(TestImgLoader)
@make_nograd_func
def do_iteration():
torch.cuda.synchronize()
torch.cuda.synchronize()
start_time = time.perf_counter()
test_sample(next(iter_dataloader), detailed_summary=True)
torch.cuda.synchronize()
end_time = time.perf_counter()
return end_time - start_time
for i in range(warmup_iter):
t = do_iteration()
print('WarpUp Iter {}, time = {:.4f}'.format(i, t))
with torch.autograd.profiler.profile(enabled=True, use_cuda=True) as prof:
for i in range(5):
t = do_iteration()
print('Profile Iter {}, time = {:.4f}'.format(i, t))
time.sleep(0.02)
if prof is not None:
# print(prof)
trace_fn = 'chrome-trace.bin'
prof.export_chrome_trace(trace_fn)
print("chrome trace file is written to: ", trace_fn)
if __name__ == '__main__':
if args.mode == "train":
train()
elif args.mode == "test":
test()
elif args.mode == "profile":
profile()