-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
404 lines (336 loc) · 16.9 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
398
399
400
401
402
403
404
import os, argparse, json, datetime
import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.backends.cudnn as cudnn
from torchvision import transforms
import numpy as np
import random
from rich.console import Console
import data as validate_data
import data_training as data
from model import load_model
def parse_args():
"""Parse input arguments."""
parser = argparse.ArgumentParser(description='Head pose estimation using the EfficientNetV2-based network.')
parser.add_argument('--flip',
dest='flip', help='flip.',
default=0, type=int)
parser.add_argument('--augment',
dest='augment', help='augment.',
default=0.5, type=float)
parser.add_argument('--gpu',
dest='gpu_id', help='GPU device id to use [0]',
default=0, type=int)
parser.add_argument('--num_epochs',
dest='num_epochs', help='Maximum number of training epochs.',
default=100, type=int)
parser.add_argument('--batch_size',
dest='batch_size', help='Batch size.',
default=32, type=int)
parser.add_argument('--lr', dest='lr',
help='Base learning rate.',
default=0.00001, type=float)
parser.add_argument('--val_data_dir',
dest='val_data_dir', help='Directory path for data.',
default='datasets/AFLW2000', type=str)
parser.add_argument('--val_filename_list', dest='val_filename_list',
help='Path to text file containing relative paths for every example.',
default='datasets/AFLW2000/aflw2000_list.txt', type=str)
parser.add_argument('--dataset', dest='dataset',
help='Dataset type.', default='300W_LP', type=str)
parser.add_argument('--val_dataset', dest='val_dataset',
help='val Dataset type.', default='AFLW2000', type=str)
parser.add_argument('--data_dir', dest='data_dir',
help='Directory path for data.', default='datasets/300W_LP', type=str)
parser.add_argument('--filename_list', dest='filename_list',
help='Path to text file containing relative paths for every example.',
default='datasets/300W_LP/300wlp_list.txt', type=str)
parser.add_argument('--target_size', dest='target_size', help='target_size',
default=224, type=int)
parser.add_argument('--transfer', dest='transfer', help='transfer.',
default=1, type=int)
parser.add_argument('--output_string', dest='output_string',
help='String appended to output snapshots.',
default = '', type=str)
parser.add_argument('--snapshot',
dest='snapshot', help='Path of model snapshot.',
default='pretrained/pretrained_s.pkl', type=str)
parser.add_argument('--efficient',
dest='efficient', help='efficient.',
default=4, type=int)
args = parser.parse_args()
return args
# Generator function that yields params that will be optimized.
def get_non_ignored_params(model):
b = [model.stem, model.blocks, model.side, model.aspp]
for i in range(len(b)):
for module_name, module in b[i].named_modules():
if 'bn' in module_name:
module.eval()
for name, param in module.named_parameters():
yield param
def get_fc_params(model):
b = [model.fc_yaw_coarse, model.fc_yaw_shift,
model.fc_pitch_coarse, model.fc_pitch_shift,
model.fc_roll_coarse, model.fc_roll_shift]
for i in range(len(b)):
for module_name, module in b[i].named_modules():
for name, param in module.named_parameters():
yield param
def wrapped_loss():
def call(predict, target):
a = torch.square(predict - target)
b = torch.square(360 - (predict - target))
c = torch.minimum(a, b)
c = torch.mean(c)
return c
return call
if __name__ == '__main__':
args = parse_args()
cudnn.enabled = True
num_epochs = args.num_epochs
batch_size = args.batch_size
gpu = args.gpu_id
seed = 0
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
np.random.seed(seed)
random.seed(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# Load Rich Console
console = Console()
nowdate = datetime.datetime.now()
nowdate_str = nowdate.strftime("%Y_%m_%d_%H_%M_%S")
if args.output_string != '':
base_path = f'workdirs_{ args.output_string }'
else:
base_path = f'workdirs'
output_path = f'{ base_path }/snapshots'
if not os.path.exists(output_path):
os.makedirs(output_path)
# EfficientNetV2 Structure
if args.efficient == 4:
network = 'efficientnet_v2_s'
elif args.efficient == 3:
network = 'efficientnet_v2_m'
else:
network = 'efficientnet_v2_l'
model = load_model(pretrained = args.snapshot == '', network = network)
pytorch_total_params = sum(p.numel() for p in model.parameters())
console.log(f"This model has [bold magenta]{ pytorch_total_params }[/bold magenta] parameters.")
optimizer = torch.optim.AdamW([
{ 'params': get_non_ignored_params(model), 'lr': args.lr },
{ 'params': get_fc_params(model), 'lr': args.lr * 5}
], lr = args.lr)
model.cuda(gpu)
if args.snapshot != '':
console.log(f"Load Snapshot from { args.snapshot }.")
saved_state_dict = torch.load(args.snapshot, map_location = 'cuda')
if args.transfer != 1:
model.load_state_dict(saved_state_dict['model'])
start_epoch = saved_state_dict['epoch']
if 'optimizer' in saved_state_dict:
optimizer.load_state_dict(saved_state_dict['optimizer'])
start_epoch += 1
else:
model.load_state_dict(saved_state_dict['model'])
start_epoch = 0
else:
start_epoch = 0
console.log('Loading data.')
if args.dataset == '300W_LP':
pose_dataset = data.Pose_300WLP_separate(args.data_dir, args.filename_list, flip = args.flip == 1,
augment = args.augment, target_size = args.target_size)
elif args.dataset == 'BIWI':
pose_dataset = data.BIWI(args.data_dir, args.filename_list, flip = args.flip == 1,
augment = args.augment, target_size = args.target_size)
elif args.dataset == 'DAD':
pose_dataset = data.Pose_DAD3DHeads(args.data_dir, args.filename_list, flip = args.flip == 1,
augment = args.augment, target_size = args.target_size)
else:
console.log('Not implement')
exit(1)
transformations = transforms.Compose([
transforms.Resize((args.target_size, args.target_size)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
if args.val_dataset =='AFLW2000':
val_pose_dataset = validate_data.AFLW2000(args.val_data_dir, args.val_filename_list, None, transformations, sixd = False, ad = 0.2)
elif args.val_dataset == 'BIWI':
val_pose_dataset = validate_data.BIWI_kinect(args.val_data_dir, args.val_filename_list, None, transformations, 'val', sixd = False)
else:
console.log('Not implement')
exit(1)
console.log(f"Training { len(pose_dataset )} images.")
# Training
train_loader = torch.utils.data.DataLoader(dataset=pose_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=4)
# Validation
test_loader = torch.utils.data.DataLoader(dataset=val_pose_dataset,
batch_size=args.batch_size,
num_workers=4)
class_criterion = nn.CrossEntropyLoss().cuda(gpu)
reg_criterion = wrapped_loss()
# Regression loss coefficient
alpha = 1
softmax = nn.Softmax().cuda(gpu)
idx_tensor = [idx for idx in range(20)]
idx_tensor = Variable(torch.FloatTensor(idx_tensor)).cuda(gpu)
ten_idx_tensor = np.array([idx for idx in range(10)])
ten_idx_tensor = Variable(torch.FloatTensor(ten_idx_tensor)).cuda(gpu)
console.log('Ready to train network.')
f = open(f"{ base_path }/training.log", "a")
for epoch in range(start_epoch, num_epochs):
model.train()
for i, (images, labels, shifted_labels, cont_labels, index) in enumerate(train_loader):
batch_size = images.size(0)
images = Variable(images).cuda(gpu)
# Rotation euler labels
label_yaw_cont = Variable(cont_labels[:,0]).cuda(gpu) #* 180 / np.pi
label_pitch_cont = Variable(cont_labels[:,1]).cuda(gpu) #* 180 / np.pi
label_roll_cont = Variable(cont_labels[:,2]).cuda(gpu) #* 180 / np.pi
# Binned labels
label_yaw = Variable(labels[:,0]).cuda(gpu)
label_pitch = Variable(labels[:,1]).cuda(gpu)
label_roll = Variable(labels[:,2]).cuda(gpu)
# Shifted Binned labels
shifted_label_yaw = Variable(shifted_labels[:,0]).cuda(gpu)
shifted_label_pitch = Variable(shifted_labels[:,1]).cuda(gpu)
shifted_label_roll = Variable(shifted_labels[:,2]).cuda(gpu)
yaw_coarse, yaw_shift, pitch_coarse, pitch_shift, roll_coarse, roll_shift = model(images)
# 20-bin loss
# Cross entropy loss
loss_yaw = class_criterion(yaw_coarse, label_yaw)
loss_pitch = class_criterion(pitch_coarse, label_pitch)
loss_roll = class_criterion(roll_coarse, label_roll)
# 20-shift bin loss
# Cross entropy loss
loss_shifted_yaw = class_criterion(yaw_shift, shifted_label_yaw)
loss_shifted_pitch = class_criterion(pitch_shift, shifted_label_pitch)
loss_shifted_roll = class_criterion(roll_shift, shifted_label_roll)
loss_yaw += loss_shifted_yaw
loss_pitch += loss_shifted_pitch
loss_roll += loss_shifted_roll
# Continuous
# Coarse
ten_yaw = softmax(yaw_coarse)
ten_pitch = softmax(pitch_coarse)
ten_roll = softmax(roll_coarse)
yaw_predicted = torch.sum(ten_yaw * ten_idx_tensor, 1) * 20 - 100
pitch_predicted = torch.sum(ten_pitch * ten_idx_tensor, 1) * 20 - 100
roll_predicted = torch.sum(ten_roll * ten_idx_tensor, 1) * 20 - 100
# Shift
shifted_yaw = softmax(yaw_shift)
shifted_pitch = softmax(pitch_shift)
shifted_roll = softmax(roll_shift)
yaw_predicted += torch.sum(shifted_yaw * idx_tensor, 1)
pitch_predicted += torch.sum(shifted_pitch * idx_tensor, 1)
roll_predicted += torch.sum(shifted_roll * idx_tensor, 1)
# MSE Loss
loss_reg_yaw = reg_criterion(yaw_predicted, label_yaw_cont)
loss_reg_pitch = reg_criterion(pitch_predicted, label_pitch_cont)
loss_reg_roll = reg_criterion(roll_predicted, label_roll_cont)
# Total loss
loss_yaw += alpha * loss_reg_yaw
loss_pitch += alpha * loss_reg_pitch
loss_roll += alpha * loss_reg_roll
# Calculation
total_loss = loss_yaw + loss_pitch + loss_roll
# Backpropagation
optimizer.zero_grad()
total_loss.backward()
optimizer.step()
if (i+1) % 50 == 0:
console.log('Epoch [%d/%d], Iter [%d/%d] Losses: Yaw %.4f Pitch %.4f Roll %.4f '
%(epoch+1, num_epochs, i+1, len(pose_dataset)//batch_size,
loss_yaw.cpu().item(), loss_pitch.cpu().item(), loss_roll.cpu().item()))
dict_loss = {
"mode": "train",
"epoch": epoch + 1,
"total_epoch": num_epochs,
"iter": i+1,
"total_iter": len(pose_dataset)//batch_size,
"loss_yaw": loss_yaw.cpu().item(),
"loss_pitch": loss_pitch.cpu().item(),
"loss_roll": loss_roll.cpu().item()
}
f.write(json.dumps(dict_loss))
f.write("\n")
# Save models at numbered epochs.
if epoch % 1 == 0 and epoch < num_epochs:
with console.status("[bold green]Taking snapshot...") as status:
checkpoint = {
'epoch': epoch,
'model': model.state_dict(),
'optimizer': optimizer.state_dict()
}
torch.save(checkpoint, f'{ output_path }/'+ str(epoch+1) + '.pkl')
model.eval()
yaw_error = .0
pitch_error = .0
roll_error = .0
with torch.no_grad():
data = { 'yaw': { "result": [], "label": [] }, 'pitch': { "result": [], "label": [] }, 'roll': { "result": [], "label": [] } }
with console.status("[bold green]Testing...") as status:
for i, test_temp in enumerate(test_loader):
if len(test_temp) == 5:
images, labels, cont_labels, raw_img, index = test_temp
else:
images, labels, cont_labels, raw_img, index = test_temp
images = Variable(images).cuda(gpu)
batch_size = cont_labels.size(0)
label_yaw = cont_labels[:,0].float()
label_pitch = cont_labels[:,1].float()
label_roll = cont_labels[:,2].float()
yaw_coarse, yaw_shift, pitch_coarse, pitch_shift, roll_coarse, roll_shift = model(images)
# Continuous
# Coarse
ten_yaw = softmax(yaw_coarse)
ten_pitch = softmax(pitch_coarse)
ten_roll = softmax(roll_coarse)
yaw_predicted = torch.sum(ten_yaw * ten_idx_tensor, 1) * 20 - 100
pitch_predicted = torch.sum(ten_pitch * ten_idx_tensor, 1) * 20 - 100
roll_predicted = torch.sum(ten_roll * ten_idx_tensor, 1) * 20 - 100
# Shift
shifted_yaw = softmax(yaw_shift)
shifted_pitch = softmax(pitch_shift)
shifted_roll = softmax(roll_shift)
yaw_predicted += torch.sum(shifted_yaw * idx_tensor, 1)
pitch_predicted += torch.sum(shifted_pitch * idx_tensor, 1)
roll_predicted += torch.sum(shifted_roll * idx_tensor, 1)
# Mean absolute error
p_gt_deg = label_pitch
y_gt_deg = label_yaw
r_gt_deg = label_roll
p_pred_deg = pitch_predicted.cpu()
y_pred_deg = yaw_predicted.cpu()
r_pred_deg = roll_predicted.cpu()
pitch_error += torch.sum(torch.min(torch.stack((torch.abs(p_gt_deg - p_pred_deg), torch.abs(p_pred_deg + 360 - p_gt_deg), torch.abs(
p_pred_deg - 360 - p_gt_deg), torch.abs(p_pred_deg + 180 - p_gt_deg), torch.abs(p_pred_deg - 180 - p_gt_deg))), 0)[0])
yaw_error += torch.sum(torch.min(torch.stack((torch.abs(y_gt_deg - y_pred_deg), torch.abs(y_pred_deg + 360 - y_gt_deg), torch.abs(
y_pred_deg - 360 - y_gt_deg), torch.abs(y_pred_deg + 180 - y_gt_deg), torch.abs(y_pred_deg - 180 - y_gt_deg))), 0)[0])
roll_error += torch.sum(torch.min(torch.stack((torch.abs(r_gt_deg - r_pred_deg), torch.abs(r_pred_deg + 360 - r_gt_deg), torch.abs(
r_pred_deg - 360 - r_gt_deg), torch.abs(r_pred_deg + 180 - r_gt_deg), torch.abs(r_pred_deg - 180 - r_gt_deg))), 0)[0])
total = len(val_pose_dataset)
mean_error = ((yaw_error + pitch_error + roll_error)) / 3
console.log('Test error in degrees of the model on the ' + str(total) +
' test images. Yaw: %.4f, Pitch: %.4f, Roll: %.4f, Mean: %.4f' % (yaw_error / total,
pitch_error / total, roll_error / total, mean_error / total))
f.write(json.dumps({
"mode": "val",
"epoch": epoch + 1,
"total_epoch": num_epochs,
"total_image": total,
"yaw_error": yaw_error.item() / total,
"pitch_error": pitch_error.item() / total,
"roll_error": roll_error.item() / total,
"mean_error": mean_error.item() / total
}))
f.write("\n")
f.close()