forked from wiibrew/pytorch-yolo2
-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
256 lines (227 loc) · 8.58 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
from __future__ import print_function
import sys
if len(sys.argv) != 4:
print('Usage:')
print('python train.py datacfg cfgfile weightfile')
exit()
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torch.backends.cudnn as cudnn
from torchvision import datasets, transforms
from torch.autograd import Variable
import dataset
import random
import math
from utils import *
from cfg import parse_cfg
from region_loss import RegionLoss
from darknet import Darknet
from models.tiny_yolo import TinyYoloNet
# Training settings
datacfg = sys.argv[1]
cfgfile = sys.argv[2]
weightfile = sys.argv[3]
data_options = read_data_cfg(datacfg)
net_options = parse_cfg(cfgfile)[0]
trainlist = data_options['train']
testlist = data_options['valid']
backupdir = data_options['backup']
nsamples = file_lines(trainlist)
gpus = data_options['gpus'] # e.g. 0,1,2,3
ngpus = len(gpus.split(','))
num_workers = int(data_options['num_workers'])
batch_size = int(net_options['batch'])
max_batches = int(net_options['max_batches'])
learning_rate = float(net_options['learning_rate'])
momentum = float(net_options['momentum'])
decay = float(net_options['decay'])
steps = [float(step) for step in net_options['steps'].split(',')]
scales = [float(scale) for scale in net_options['scales'].split(',')]
#Train parameters
max_epochs = max_batches*batch_size/nsamples+1
use_cuda = True
seed = int(time.time())
eps = 1e-5
save_interval = 10 # epoches
dot_interval = 70 # batches
# Test parameters
conf_thresh = 0.25
nms_thresh = 0.4
iou_thresh = 0.5
###############
torch.manual_seed(seed)
if use_cuda:
os.environ['CUDA_VISIBLE_DEVICES'] = gpus
torch.cuda.manual_seed(seed)
model = Darknet(cfgfile)
region_loss = model.loss
model.load_weights(weightfile)
model.print_network()
region_loss.seen = model.seen
processed_batches = model.seen/batch_size
init_width = model.width
init_height = model.height
init_epoch = model.seen/nsamples
kwargs = {'num_workers': num_workers, 'pin_memory': True} if use_cuda else {}
test_loader = torch.utils.data.DataLoader(
dataset.listDataset(testlist, shape=(init_width, init_height),
shuffle=False,
transform=transforms.Compose([
transforms.ToTensor(),
]), train=False),
batch_size=batch_size, shuffle=False, **kwargs)
if use_cuda:
if ngpus > 1:
model = torch.nn.DataParallel(model).cuda()
else:
model = model.cuda()
params_dict = dict(model.named_parameters())
params = []
for key, value in params_dict.items():
if key.find('.bn') >= 0 or key.find('.bias') >= 0:
params += [{'params': [value], 'weight_decay': 0.0}]
else:
params += [{'params': [value], 'weight_decay': decay*batch_size}]
optimizer = optim.SGD(model.parameters(), lr=learning_rate/batch_size, momentum=momentum, dampening=0, weight_decay=decay*batch_size)
def adjust_learning_rate(optimizer, batch):
"""Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
lr = learning_rate
for i in range(len(steps)):
scale = scales[i] if i < len(scales) else 1
if batch >= steps[i]:
lr = lr * scale
if batch == steps[i]:
break
else:
break
for param_group in optimizer.param_groups:
param_group['lr'] = lr/batch_size
return lr
def train(epoch):
global processed_batches
t0 = time.time()
if ngpus > 1:
cur_model = model.module
else:
cur_model = model
train_loader = torch.utils.data.DataLoader(
dataset.listDataset(trainlist, shape=(init_width, init_height),
shuffle=True,
transform=transforms.Compose([
transforms.ToTensor(),
]),
train=True,
seen=cur_model.seen,
batch_size=batch_size,
num_workers=num_workers),
batch_size=batch_size, shuffle=False, **kwargs)
lr = adjust_learning_rate(optimizer, processed_batches)
logging('epoch %d, processed %d samples, lr %f' % (epoch, epoch * len(train_loader.dataset), lr))
model.train()
t1 = time.time()
avg_time = torch.zeros(9)
for batch_idx, (data, target) in enumerate(train_loader):
t2 = time.time()
adjust_learning_rate(optimizer, processed_batches)
processed_batches = processed_batches + 1
#if (batch_idx+1) % dot_interval == 0:
# sys.stdout.write('.')
if use_cuda:
data = data.cuda()
#target= target.cuda()
t3 = time.time()
data, target = Variable(data), Variable(target)
t4 = time.time()
optimizer.zero_grad()
t5 = time.time()
output = model(data)
t6 = time.time()
region_loss.seen = region_loss.seen + data.data.size(0)
loss = region_loss(output, target)
t7 = time.time()
loss.backward()
t8 = time.time()
optimizer.step()
t9 = time.time()
if False and batch_idx > 1:
avg_time[0] = avg_time[0] + (t2-t1)
avg_time[1] = avg_time[1] + (t3-t2)
avg_time[2] = avg_time[2] + (t4-t3)
avg_time[3] = avg_time[3] + (t5-t4)
avg_time[4] = avg_time[4] + (t6-t5)
avg_time[5] = avg_time[5] + (t7-t6)
avg_time[6] = avg_time[6] + (t8-t7)
avg_time[7] = avg_time[7] + (t9-t8)
avg_time[8] = avg_time[8] + (t9-t1)
print('-------------------------------')
print(' load data : %f' % (avg_time[0]/(batch_idx)))
print(' cpu to cuda : %f' % (avg_time[1]/(batch_idx)))
print('cuda to variable : %f' % (avg_time[2]/(batch_idx)))
print(' zero_grad : %f' % (avg_time[3]/(batch_idx)))
print(' forward feature : %f' % (avg_time[4]/(batch_idx)))
print(' forward loss : %f' % (avg_time[5]/(batch_idx)))
print(' backward : %f' % (avg_time[6]/(batch_idx)))
print(' step : %f' % (avg_time[7]/(batch_idx)))
print(' total : %f' % (avg_time[8]/(batch_idx)))
t1 = time.time()
print('')
t1 = time.time()
logging('training with %f samples/s' % (len(train_loader.dataset)/(t1-t0)))
if (epoch+1) % save_interval == 0:
logging('save weights to %s/%06d.weights' % (backupdir, epoch+1))
cur_model.seen = (epoch + 1) * len(train_loader.dataset)
cur_model.save_weights('%s/%06d.weights' % (backupdir, epoch+1))
def test(epoch):
def truths_length(truths):
for i in range(50):
if truths[i][1] == 0:
return i
model.eval()
if ngpus > 1:
cur_model = model.module
else:
cur_model = model
num_classes = cur_model.num_classes
anchors = cur_model.anchors
num_anchors = cur_model.num_anchors
total = 0.0
proposals = 0.0
correct = 0.0
for batch_idx, (data, target) in enumerate(test_loader):
if use_cuda:
data = data.cuda()
data = Variable(data, volatile=True)
output = model(data).data
all_boxes = get_region_boxes(output, conf_thresh, num_classes, anchors, num_anchors)
for i in range(output.size(0)):
boxes = all_boxes[i]
boxes = nms(boxes, nms_thresh)
truths = target[i].view(-1, 5)
num_gts = truths_length(truths)
total = total + num_gts
for i in range(len(boxes)):
if boxes[i][4] > conf_thresh:
proposals = proposals+1
for i in range(num_gts):
box_gt = [truths[i][1], truths[i][2], truths[i][3], truths[i][4], 1.0, 1.0, truths[i][0]]
best_iou = 0
for j in range(len(boxes)):
iou = bbox_iou(box_gt, boxes[j], x1y1x2y2=False)
best_iou = max(iou, best_iou)
if best_iou > iou_thresh and boxes[j][6] == box_gt[6]:
correct = correct+1
precision = 1.0*correct/(proposals+eps)
recall = 1.0*correct/(total+eps)
fscore = 2.0*precision*recall/(precision+recall+eps)
logging("precision: %f, recall: %f, fscore: %f" % (precision, recall, fscore))
evaluate = False
if evaluate:
logging('evaluating ...')
test(0)
else:
for epoch in range(init_epoch, max_epochs):
train(epoch)
test(epoch)