-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathtrain.py
200 lines (152 loc) · 7.13 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
import argparse
import collections
import gc
import numpy as np
import time
import torch
import torch.optim as optim
from torchvision import transforms
from torch.profiler import profile, record_function, ProfilerActivity
from data_processing.generator.crf_main_generator import create_generators
from utils.config import get_config
from model.architecture.retinanet import Retinanet
from model.architecture.vgg import Vggmax
from data_processing.dataloader import CocoDataset, CSVDataset, collater, Resizer, AspectRatioBasedSampler, Augmenter, \
Normalizer
from torch.utils.data import DataLoader
from model import nus_eval
assert torch.__version__.split('.')[0] == '1'
print('CUDA available: {}'.format(torch.cuda.is_available()))
def main(args=None):
parser = argparse.ArgumentParser(description='Simple training script for training a RetinaNet network.')
parser.add_argument('--epochs', help='Number of epochs', type=int, default=100)
parser.add_argument('--radar', help='Use radar modality?', type=bool, default=False)
parser.add_argument('--load_path', help='Load model path', type=str, default=None)
parser = parser.parse_args(args)
# Create the data loaders
###### New DataLoaders here
backbone = Vggmax(parser.radar)
cfg = get_config('./config/default.cfg')
train_generator, validation_generator, test_generator, test_night_generator, test_rain_generator = create_generators(cfg, backbone)
dataloader = DataLoader(train_generator, batch_size=cfg.batchsize, num_workers=10)
batch_size = cfg.batchsize
print('=='*45)
print('Length of train set: {}, val set: {}, test set: {}, test_night set: {}, test_rain set: {}'.format(len(train_generator), len(validation_generator), len(test_generator), len(test_night_generator), len(test_rain_generator)))
print('=='*45)
#####
#print(train_generator[0][0][:,:,:,0])
#exit(0)
# Create the model
image_size = (360, 640)
if parser.radar:
f_size = 254
else:
f_size = 256
retinanet = Retinanet(backbone, pretrained=True, num_anchors=9, num_classes=train_generator.num_classes(), feature_size=f_size, image_size=image_size)
if parser.load_path is not None:
retinanet.load_state_dict(torch.load(parser.load_path, map_location='cuda:0'))
start_ep = 20
else:
start_ep = 0
use_gpu = True
if use_gpu:
if torch.cuda.is_available():
retinanet = retinanet.cuda()
if torch.cuda.is_available():
retinanet = torch.nn.DataParallel(retinanet).cuda()
else:
retinanet = torch.nn.DataParallel(retinanet)
retinanet.training = True
optimizer = optim.Adam(retinanet.parameters(), lr=1e-5)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer, patience=3, verbose=True)
loss_hist = collections.deque(maxlen=500)
#retinanet.train()
#retinanet.module.freeze_bn()
print('===='*7 + '\nNum training images: {}\n'.format(len(train_generator)) + '===='*10)
start_time = time.time()
for epoch_num in range(start_ep, parser.epochs):
retinanet.train()
#retinanet.module.freeze_bn()
epoch_loss = []
#for iter_num, data in enumerate(train_generator):
for iter_num, data in enumerate(dataloader):#range(len(train_generator)):
#try:
#with profile(activities=[
# ProfilerActivity.CPU, ProfilerActivity.CUDA], record_shapes=True) as prof:
# with record_function("model_inference"):
optimizer.zero_grad()
#print('Data shape: {}, Annotation shape: {}'.format(data[0].shape, data[1][0].shape))
#Data format data[0] = 5 Channel last image, data[1][0] = Regression annot, data[1][1] = classification annot.
#data = train_generator[iter_num]
#print(data[0].shape)
#print(len(data[1]))
#print(data[1][0].shape)
#print(data[1][1].shape)
if not parser.radar: # Crop and keep only image channels
img = data[0][0,:,:,:,:3]
else:
img = data[0][0,:,:,:,:]
img = torch.permute(img.cuda().float(), (0,3,1,2))
targets = get_annotations_for_batch(train_generator, iter_num, batch_size)
#print('==='*10)
#print(img[0,0,:,:])
#print(torch.max(img[0,0,:,:]))
#print('==='*10)
#print(img[0,3,:,:])
#print(torch.max(img[0,3,:,:]))
#print('==='*10)
#print(targets[0][:,:])
#print(torch.max(targets[0][:,:]))
#print('==='*10)
#exit(0)
'''
ann = train_generator.load_annotations(iter_num)
if (len(ann['labels']) == 0):
targets = torch.tensor([[[-1, -1, -1, -1, -1]]]).cuda()
else:
#print(ann['bboxes'].shape, ann['labels'].shape)
targets = np.hstack((ann['bboxes'], np.expand_dims(ann['labels'], axis=1)))
targets = torch.tensor(np.expand_dims(targets, axis=0)).cuda()
'''
classification_loss, regression_loss = retinanet([img, targets])
classification_loss = classification_loss.mean()
regression_loss = regression_loss.mean()
loss = classification_loss + regression_loss
if bool(loss == 0):
continue
loss.backward()
torch.nn.utils.clip_grad_norm_(retinanet.parameters(), 0.1)
optimizer.step()
loss_hist.append(float(loss))
epoch_loss.append(float(loss))
print(
'Ep: {} | Iter: {} | Cls loss: {:1.5f} | Reg loss: {:1.5f} | Running loss: {:1.5f} | Elp Time: {}'.format(epoch_num, iter_num, float(classification_loss), float(regression_loss), np.mean(loss_hist), int(time.time()-start_time)), end='\r')
#except Exception as e:
# print(e)
# continue
del classification_loss, regression_loss
#print(prof.key_averages().table(sort_by="cuda_time_total", row_limit=10))
torch.save(retinanet.module.state_dict(), 'exp4_radar_image_retinanet_{}.pt'.format(epoch_num))
#try:
# mAP = nus_eval.evaluate(train_generator, retinanet)
#except Exception as e:
# print(e)
# continue
scheduler.step(np.mean(epoch_loss))
retinanet.eval()
torch.save(retinanet, 'exp2_radar_image_model_final.pt')
def get_annotations_for_batch(generator, iter_num, batch_size):
bstart = iter_num*batch_size
btargets = []
for i in range(batch_size):
ann = generator.load_annotations(bstart+i)
if (len(ann['labels']) == 0):
btargets.append(torch.tensor([[-1., -1., -1., -1., -1.]]).cuda())
else:
#print(ann['bboxes'].shape, ann['labels'].shape)
targets = np.hstack((ann['bboxes'], np.expand_dims(ann['labels'], axis=1)))
btargets.append(torch.tensor(targets).cuda())
#targets = torch.tensor(np.expand_dims(targets, axis=0)).cuda()
return btargets
if __name__ == '__main__':
main()