-
Notifications
You must be signed in to change notification settings - Fork 13
/
lt_c_train.py
executable file
·333 lines (295 loc) · 16.5 KB
/
lt_c_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
r"""PyTorch Detection Training.
To run in a multi-gpu environment, use the distributed launcher::
python -m torch.distributed.launch --nproc_per_node=$NGPU --use_env \
train.py ... --world-size $NGPU
The default hyperparameters are tuned for training on 8 gpus and 2 images per gpu.
--lr 0.02 --batch-size 2 --world-size 8
If you use different number of gpus, the learning rate should be changed to 0.02/8*$NGPU.
On top of that, for training Faster/Mask R-CNN, the default hyperparameters are
--epochs 26 --lr-steps 16 22 --aspect-ratio-group-factor 3
Also, if you train Keypoint R-CNN, the default hyperparameters are
--epochs 46 --lr-steps 36 43 --aspect-ratio-group-factor 3
Because the number of images is smaller in the person keypoint subset of COCO,
the number of epochs should be adapted so that we have the same number of iterations.
"""
import datetime
import os
import time
import random
import math
import sys
import numpy as np
import math
import torch
import torch.utils.data
from torch import nn
import torchvision
import torchvision.models.detection
import torchvision.models.detection.mask_rcnn
import torch.optim as optim
from torch.utils.data import DataLoader
import torch.optim.lr_scheduler as lr_scheduler
from torch.utils.data.sampler import SubsetRandomSampler
from detection.frcnn_la import fasterrcnn_resnet50_fpn_feature
from detection.coco_utils import get_coco, get_coco_kp
from detection.group_by_aspect_ratio import GroupedBatchSampler, create_aspect_ratio_groups
from detection.engine import coco_evaluate, voc_evaluate
from detection import utils
from detection import transforms as T
from detection.train import *
from ll4al.data.sampler import SubsetSequentialSampler
import pickle
def train_one_epoch(task_model, task_optimizer, data_loader, device, cycle, epoch, print_freq):
task_model.train()
metric_logger = utils.MetricLogger(delimiter=" ")
metric_logger.add_meter('task_lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
header = 'Cycle:[{}] Epoch: [{}]'.format(cycle, epoch)
task_lr_scheduler = None
if epoch == 0:
warmup_factor = 1. / 1000
warmup_iters = min(1000, len(data_loader) - 1)
task_lr_scheduler = utils.warmup_lr_scheduler(task_optimizer, warmup_iters, warmup_factor)
for images, targets in metric_logger.log_every(data_loader, print_freq, header):
images = list(image.to(device) for image in images)
targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
task_loss_dict = task_model(images, targets)
task_losses = sum(loss for loss in task_loss_dict.values())
# reduce losses over all GPUs for logging purposes
task_loss_dict_reduced = utils.reduce_dict(task_loss_dict)
task_losses_reduced = sum(loss.cpu() for loss in task_loss_dict_reduced.values())
task_loss_value = task_losses_reduced.item()
if not math.isfinite(task_loss_value):
print("Loss is {}, stopping training".format(task_loss_value))
print(task_loss_dict_reduced)
sys.exit(1)
task_optimizer.zero_grad()
task_losses.backward()
task_optimizer.step()
if task_lr_scheduler is not None:
task_lr_scheduler.step()
metric_logger.update(task_loss=task_losses_reduced)
metric_logger.update(task_lr=task_optimizer.param_groups[0]["lr"])
return metric_logger
def calcu_iou(A, B):
'''
calculate two box's iou
'''
width = min(A[2], B[2]) - max(A[0], B[0]) + 1
height = min(A[3], B[3]) - max(A[1], B[1]) + 1
if width <= 0 or height <= 0:
return 0
Aarea = (A[2] - A[0]) * (A[3] - A[1] + 1)
Barea = (B[2] - B[0]) * (B[3] - B[1] + 1)
iner_area = width * height
return iner_area / (Aarea + Barea - iner_area)
def get_uncertainty(task_model, unlabeled_loader):
task_model.eval()
uncertainties = []
with torch.no_grad():
for images, labels in unlabeled_loader:
images = list(img.cuda() for img in images)
torch.cuda.synchronize()
outputs = task_model(images)
for output in outputs:
uncertainty = 1.0
for box, prop, prob_max in zip(output['boxes'], output['props'], output['prob_max']):
iou = calcu_iou(box, prop)
u = torch.abs(iou + prob_max - 1)
uncertainty = min(uncertainty, u.item())
uncertainties.append(uncertainty)
return uncertainties
def main(args):
torch.cuda.set_device(0)
random.seed(0)
torch.manual_seed(0)
torch.cuda.manual_seed(0)
torch.cuda.manual_seed_all(0)
print(args)
device = torch.device(args.device)
# Data loading code
print("Loading data")
if '2007' in args.dataset:
dataset, num_classes = get_dataset(args.dataset, "trainval", get_transform(train=True), args.data_path)
dataset_test, _ = get_dataset(args.dataset, "test", get_transform(train=False), args.data_path)
else:
dataset, num_classes = get_dataset(args.dataset, "train", get_transform(train=True), args.data_path)
dataset_test, _ = get_dataset(args.dataset, "val", get_transform(train=False), args.data_path)
print("Creating data loaders")
num_images = len(dataset)
if 'voc' in args.dataset:
init_num = 500
budget_num = 500
else:
init_num = 5000
budget_num = 1000
indices = list(range(num_images))
random.shuffle(indices)
labeled_set = indices[:init_num]
unlabeled_set = indices[init_num:]
train_sampler = SubsetRandomSampler(labeled_set)
test_sampler = torch.utils.data.SequentialSampler(dataset_test)
data_loader_test = DataLoader(dataset_test, batch_size=1, sampler=test_sampler, num_workers=args.workers,
collate_fn=utils.collate_fn)
for cycle in range(args.cycles):
if args.aspect_ratio_group_factor >= 0:
group_ids = create_aspect_ratio_groups(dataset, k=args.aspect_ratio_group_factor)
train_batch_sampler = GroupedBatchSampler(train_sampler, group_ids, args.batch_size)
else:
train_batch_sampler = torch.utils.data.BatchSampler(train_sampler, args.batch_size, drop_last=True)
data_loader = torch.utils.data.DataLoader(dataset, batch_sampler=train_batch_sampler, num_workers=args.workers,
collate_fn=utils.collate_fn)
print("Creating model")
if 'voc' in args.dataset:
task_model = fasterrcnn_resnet50_fpn_feature(num_classes=num_classes, min_size=600, max_size=1000)
else:
task_model = fasterrcnn_resnet50_fpn_feature(num_classes=num_classes, min_size=800, max_size=1333)
task_model.to(device)
if not args.init and cycle == 0 and args.skip:
checkpoint = torch.load(os.path.join(args.first_checkpoint_path, '{}_frcnn_1st.pth'.format(args.dataset)),
map_location='cpu')
task_model.load_state_dict(checkpoint['model'])
# if 'coco' in args.dataset:
# coco_evaluate(task_model, data_loader_test)
# elif 'voc' in args.dataset:
# voc_evaluate(task_model, data_loader_test, args.dataset)
print("Getting stability")
random.shuffle(unlabeled_set)
if 'coco' in args.dataset:
subset = unlabeled_set[:10000]
else:
subset = unlabeled_set
labeled_loader = DataLoader(dataset, batch_size=1,
sampler=SubsetSequentialSampler(labeled_set), num_workers=args.workers,
# more convenient if we maintain the order of subset
pin_memory=True, collate_fn=utils.collate_fn)
u = get_uncertainty(task_model, labeled_loader)
with open("vis/ltc_labeled_metric_{}_{}_{}.pkl".format(args.model, args.dataset, cycle),
"wb") as fp: # Pickling
pickle.dump(u, fp)
unlabeled_loader = DataLoader(dataset, batch_size=1, sampler=SubsetSequentialSampler(subset),
num_workers=args.workers, pin_memory=True, collate_fn=utils.collate_fn)
uncertainty = get_uncertainty(task_model, unlabeled_loader)
arg = np.argsort(uncertainty)
with open("vis/ltc_unlabeled_metric_{}_{}_{}.pkl".format(args.model, args.dataset, cycle),
"wb") as fp: # Pickling
pickle.dump(torch.tensor(uncertainty)[arg][:budget_num].numpy(), fp)
# Update the labeled dataset and the unlabeled dataset, respectively
labeled_set += list(torch.tensor(subset)[arg][:budget_num].numpy())
# file = open('vis/lt_c_{}.pkl'.format(cycle), "wb")
# pickle.dump(labeled_set, file) # 保存list到文件
# file.close()
unlabeled_set = list(set(indices) - set(labeled_set))
# Create a new dataloader for the updated labeled dataset
train_sampler = SubsetRandomSampler(labeled_set)
continue
params = [p for p in task_model.parameters() if p.requires_grad]
task_optimizer = torch.optim.SGD(params, lr=args.lr, momentum=args.momentum, weight_decay=args.weight_decay)
task_lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(task_optimizer, milestones=args.lr_steps,
gamma=args.lr_gamma)
# Start active learning cycles training
if args.test_only:
if 'coco' in args.dataset:
coco_evaluate(task_model, data_loader_test)
elif 'voc' in args.dataset:
voc_evaluate(task_model, data_loader_test, args.dataset, path=args.results_path)
return
print("Start training")
start_time = time.time()
for epoch in range(args.start_epoch, args.total_epochs):
train_one_epoch(task_model, task_optimizer, data_loader, device, cycle, epoch, args.print_freq)
task_lr_scheduler.step()
# evaluate after pre-set epoch
if (epoch + 1) == args.total_epochs:
if 'coco' in args.dataset:
coco_evaluate(task_model, data_loader_test)
elif 'voc' in args.dataset:
voc_evaluate(task_model, data_loader_test, args.dataset)
if not args.skip and cycle == 0:
utils.save_on_master({
'model': task_model.state_dict(), 'args': args},
os.path.join(args.first_checkpoint_path, '{}_frcnn_1st.pth'.format(args.dataset)))
random.shuffle(unlabeled_set)
if 'coco' in args.dataset:
subset = unlabeled_set[:10000]
else:
subset = unlabeled_set
labeled_loader = DataLoader(dataset, batch_size=1,
sampler=SubsetSequentialSampler(labeled_set), num_workers=args.workers,
# more convenient if we maintain the order of subset
pin_memory=True, collate_fn=utils.collate_fn)
u = get_uncertainty(task_model, labeled_loader)
with open("vis/ltc_labeled_metric_{}_{}_{}.pkl".format(args.model, args.dataset, cycle),
"wb") as fp: # Pickling
pickle.dump(u, fp)
unlabeled_loader = DataLoader(dataset, batch_size=1, sampler=SubsetSequentialSampler(subset),
num_workers=args.workers, pin_memory=True, collate_fn=utils.collate_fn)
uncertainty = get_uncertainty(task_model, unlabeled_loader)
arg = np.argsort(uncertainty)
with open("vis/ltc_unlabeled_metric_{}_{}_{}.pkl".format(args.model, args.dataset, cycle),
"wb") as fp: # Pickling
pickle.dump(torch.tensor(uncertainty)[arg][:budget_num].numpy(), fp)
# Update the labeled dataset and the unlabeled dataset, respectively
labeled_set += list(torch.tensor(subset)[arg][:budget_num].numpy())
# file = open('vis/lt_c_{}.pkl'.format(cycle), "wb")
# pickle.dump(labeled_set, file) # 保存list到文件
# file.close()
unlabeled_set = list(set(indices) - set(labeled_set))
# Create a new dataloader for the updated labeled dataset
train_sampler = SubsetRandomSampler(labeled_set)
total_time = time.time() - start_time
total_time_str = str(datetime.timedelta(seconds=int(total_time)))
print('Training time {}'.format(total_time_str))
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser(
description=__doc__)
parser.add_argument('-p', '--data-path', default='/data/yuweiping/voc/', help='dataset path')
parser.add_argument('--dataset', default='voc2012', help='dataset')
parser.add_argument('--model', default='fasterrcnn_resnet50_fpn', help='model')
parser.add_argument('--device', default='cuda', help='device')
parser.add_argument('-b', '--batch-size', default=4, type=int,
help='images per gpu, the total batch size is $NGPU x batch_size')
parser.add_argument('-cp', '--first-checkpoint-path', default='/data/yuweiping/voc/',
help='path to save checkpoint of first cycle')
parser.add_argument('--task_epochs', default=20, type=int, metavar='N',
help='number of total epochs to run')
parser.add_argument('-e', '--total_epochs', default=26, type=int, metavar='N',
help='number of total epochs to run')
parser.add_argument('--cycles', default=7, type=int, metavar='N',
help='number of cycles epochs to run')
parser.add_argument('-j', '--workers', default=4, type=int, metavar='N',
help='number of data loading workers (default: 4)')
parser.add_argument('--lr', default=0.0025, type=float,
help='initial learning rate, 0.02 is the default value for training '
'on 8 gpus and 2 images_per_gpu')
parser.add_argument('--ll-weight', default=0.5, type=float,
help='ll loss weight')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
help='momentum')
parser.add_argument('--wd', '--weight-decay', default=1e-4, type=float,
metavar='W', help='weight decay (default: 1e-4)',
dest='weight_decay')
parser.add_argument('--lr-step-size', default=8, type=int, help='decrease lr every step-size epochs')
parser.add_argument('--lr-steps', default=[16, 22], nargs='+', type=int, help='decrease lr every step-size epochs')
parser.add_argument('--lr-gamma', default=0.1, type=float, help='decrease lr by a factor of lr-gamma')
parser.add_argument('--print-freq', default=1000, type=int, help='print frequency')
parser.add_argument('--output-dir', default=None, help='path where to save')
parser.add_argument('--resume', default='', help='resume from checkpoint')
parser.add_argument('-rp', '--results-path', default='results',
help='path to save detection results (only for voc)')
parser.add_argument('--start_epoch', default=0, type=int, help='start epoch')
parser.add_argument('--aspect-ratio-group-factor', default=3, type=int)
parser.add_argument('-i', "--init", dest="init", help="if use init sample", action="store_true")
parser.add_argument("--test-only", dest="test_only", help="Only test the model", action="store_true")
parser.add_argument('-s', "--skip", dest="skip", help="Skip first cycle and use pretrained model to save time",
action="store_true")
parser.add_argument("--pretrained", dest="pretrained", help="Use pre-trained models from the modelzoo",
action="store_true")
# distributed training parameters
parser.add_argument('--world-size', default=1, type=int,
help='number of distributed processes')
parser.add_argument('--dist-url', default='env://', help='url used to set up distributed training')
args = parser.parse_args()
if args.output_dir:
utils.mkdir(args.output_dir)
main(args)