-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrainer.py
261 lines (229 loc) · 12.4 KB
/
trainer.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
import torch
import time, os
import numpy as np
from torchvision.utils import make_grid
from torchvision import transforms
from utils import transforms as local_transforms
from base import BaseTrainer, DataPrefetcher
from utils.helpers import colorize_mask
from utils.metrics import eval_metrics, AverageMeter
import matplotlib.pyplot as plt
from tqdm import tqdm
from PIL import Image
from sklearn.metrics import confusion_matrix, plot_confusion_matrix
def rgb2yuv(x):
'''convert batched rgb tensor to yuv'''
out = x.clone()
out[:, 0, :, :] = 0.299 * x[:, 0, :, :] + 0.587 * x[:, 1, :, :] + 0.114 * x[:, 2, :, :]
out[:, 1, :, :] = -0.168736 * x[:, 0, :, :] - 0.331264 * x[:, 1, :, :] + 0.5 * x[:, 2, :, :]
out[:, 2, :, :] = 0.5 * x[:, 0, :, :] - 0.418688 * x[:, 1, :, :] - 0.081312 * x[:, 2, :, :]
return out
class Trainer(BaseTrainer):
def __init__(self, model, loss, resume, config, train_loader, val_loader=None, train_logger=None, prefetch=True):
super(Trainer, self).__init__(model, loss, resume, config, train_loader, val_loader, train_logger)
self.wrt_mode, self.wrt_step = 'train_', 0
# self.log_step = config['trainer'].get('log_per_iter', int(np.sqrt(self.train_loader.batch_size)))
self.log_step = config.log_per_iter
if config.log_per_iter: self.log_step = int(self.log_step / self.train_loader.batch_size) + 1
self.num_classes = config.classes
# self.lossn = config.loss
# TRANSORMS FOR VISUALIZATION
self.restore_transform = transforms.Compose([
local_transforms.DeNormalize(self.train_loader.dataset.MEAN, self.train_loader.dataset.STD),##self.train_loader.MEAN, self.train_loader.STD),
transforms.ToPILImage()])
self.viz_transform = transforms.Compose([
transforms.Resize((400, 400)),
transforms.ToTensor()])
if self.device == torch.device('cpu'): prefetch = False
if prefetch:
self.train_loader = DataPrefetcher(train_loader, device=self.device)
self.val_loader = DataPrefetcher(val_loader, device=self.device)
torch.backends.cudnn.benchmark = True
def _train_epoch(self, epoch):
# self.logger.info('\n')
self.model.train()
if self.config.freeze_bn:
if isinstance(self.model, torch.nn.DataParallel): self.model.module.freeze_bn()
else: self.model.freeze_bn()
self.wrt_mode = 'train'
palette = self.train_loader.dataset.palette
tic = time.time()
self._reset_metrics()
# tbar = tqdm(self.train_loader, ncols=130)
val_visual = []
for batch_idx, (data, target) in enumerate(self.train_loader):
self.data_time.update(time.time() - tic)
data, target = data.to(self.device), target.to(self.device)
# print('check load train loader', target.size())
if target.size()[-1]==3:
target, backgrd = target[:, :, :, 0].clone(), target[:, :, :, 1].clone()###三通道色图
backgrd[backgrd == 255] = 1###binarize image, 防止loss过大
else:
backgrd = None
# print('check backgrd ', backgrd)
self.lr_scheduler.step(epoch=epoch-1)
if self.config.background:####外框式使用前背景
bbkg = backgrd.unsqueeze(dim=1).repeat(1, 3, 1, 1)
print('check size ', data.size(), backgrd.size(), bbkg.size())
data = data*bbkg
# LOSS & OPTIMIZE
self.optimizer.zero_grad()
if self.config.addhsv and self.config.resume:####稳定学习增强，废止
data1 = rgb2yuv(data)
output = self.model(data1)
else:
output = self.model(data)
# print('check label max ', torch.max(torch.max(target[0])))
# print('check model input output', output[0].size(), data[0].size(), target.size())
if self.config.arch[:3] == 'PSP':
assert output[0].size()[2:] == target.size()[1:]
assert output[0].size()[1] == self.num_classes
loss = self.loss(output[0], target, backgrd)
loss = loss + self.loss(output[1], target, backgrd) * 0.4###多层loss
output = output[0]
else:
assert output.size()[2:] == target.size()[1:]
assert output.size()[1] == self.num_classes
loss = self.loss(output, target, backgrd)
if isinstance(self.loss, torch.nn.DataParallel):
loss = loss.mean()
loss.backward()
self.optimizer.step()
self.total_loss.update(loss.item())
# measure elapsed time
self.batch_time.update(time.time() - tic)
tic = time.time()
# FOR EVAL
seg_metrics = eval_metrics(output, target, self.num_classes)
self._update_seg_metrics(*seg_metrics)
pixAcc, mIoU, _ = self._get_seg_metrics().values()
if batch_idx % int(len(self.train_loader) - 1) == 0 and batch_idx > 0 and epoch % 5 == 0:
print('TRAIN ({}) | Loss: {:.3f} | Acc {:.2f} mIoU {:.2f} | B {:.2f} D {:.2f} |'.format(
epoch, self.total_loss.average, pixAcc, mIoU, self.batch_time.average, self.data_time.average))
# LOGGING & TENSORBOARD
if self.config.tensorboard:
if batch_idx % int(len(self.train_loader) - 1) == 0 and batch_idx > 0 and epoch % 5 == 0:
self.wrt_step = (epoch - 1) * len(self.train_loader) + batch_idx
self.writer.add_scalar(f'{self.wrt_mode}/loss', loss.item(), self.wrt_step)
# print('TRAIN ({}) | Loss: {:.3f} | Acc {:.2f} mIoU {:.2f} | B {:.2f} D {:.2f} |'.format(
# epoch, self.total_loss.average, pixAcc, mIoU, self.batch_time.average, self.data_time.average))
# LIST OF IMAGE TO VIZ (15 images)
if len(val_visual) < 10:###如果记录文件大，减小此值
target_np = target.data.cpu().numpy()
output_np = output.data.max(1)[1].cpu().numpy()
val_visual.append([data[0].data.cpu(), target_np[0], output_np[0]])
# METRICS TO TENSORBOARD
seg_metrics = self._get_seg_metrics()
if self.config.tensorboard:
for k, v in list(seg_metrics.items())[:-1]:
self.writer.add_scalar(f'{self.wrt_mode}/{k}', v, self.wrt_step)
for i, opt_group in enumerate(self.optimizer.param_groups):
self.writer.add_scalar(f'{self.wrt_mode}/Learning_rate_{i}', opt_group['lr'], self.wrt_step)
#self.writer.add_scalar(f'{self.wrt_mode}/Momentum_{k}', opt_group['momentum'], self.wrt_step)
# WRTING & VISUALIZING THE MASKS
val_img = []
for d, t, o in val_visual:
d = self.restore_transform(d)
t, o = colorize_mask(t, palette), colorize_mask(o, palette)
d, t, o = d.convert('RGB'), t.convert('RGB'), o.convert('RGB')
[d, t, o] = [self.viz_transform(x) for x in [d, t, o]]
val_img.extend([d, t, o])
val_img = torch.stack(val_img, 0)
val_img = make_grid(val_img.cpu(), nrow=3, padding=5)
####保存图像
if epoch > 270 and epoch % 50 == 0:
# self.writer.add_image(f'{self.wrt_mode}/inputs_targets_predictions', val_img, self.wrt_step)
ndarr = val_img.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to('cpu', torch.uint8).numpy()
im = Image.fromarray(ndarr)
im.save(os.path.join(self.config.log_dir, str(epoch) + self.config.loss + 'train.png'))
# RETURN LOSS & METRICS
log = {'loss': self.total_loss.average, **seg_metrics}
#if self.lr_scheduler is not None: self.lr_scheduler.step()
return log
def _valid_epoch(self, epoch):
if self.val_loader is None:
self.logger.warning('Not data loader was passed for the validation step, No validation is performed !')
return {}
self.model.eval()
self.wrt_mode = 'val'
self._reset_metrics()
# tbar = tqdm(self.val_loader, ncols=130)
with torch.no_grad():
val_visual = []
for batch_idx, (data, target) in enumerate(self.val_loader):
#data, target = data.to(self.device), target.to(self.device)
if target.size()[-1] == 3:
target, backgrd = target[:, :, :, 0].clone(), target[:, :, :, 1].clone()
backgrd[backgrd == 255] = 1 ###binary image, 防止loss过大
else:
backgrd = None
if self.config.background:
data = data * backgrd
# LOSS
output = self.model(data)
loss = self.loss(output, target, backgrd)
if isinstance(self.loss, torch.nn.DataParallel):
loss = loss.mean()
self.total_loss.update(loss.item())
seg_metrics = eval_metrics(output, target, self.num_classes)
self._update_seg_metrics(*seg_metrics)
# LIST OF IMAGE TO VIZ (15 images)
if len(val_visual) < 10:
target_np = target.data.cpu().numpy()
output_np = output.data.max(1)[1].cpu().numpy()
val_visual.append([data[0].data.cpu(), target_np[0], output_np[0]])
# PRINT INFO
pixAcc, mIoU, _ = self._get_seg_metrics().values()
if batch_idx % int(len(self.val_loader) -1) == 0 and batch_idx > 0 and epoch % 5 == 0:
print('EVAL ({}) | Loss: {:.3f}, PixelAcc: {:.2f}, Mean IoU: {:.2f} |'.format(epoch,
self.total_loss.average, pixAcc, mIoU))
seg_metrics = self._get_seg_metrics()
# WRTING & VISUALIZING THE MASKS
val_img = []
palette = self.train_loader.dataset.palette
for d, t, o in val_visual:
d = self.restore_transform(d)
t, o = colorize_mask(t, palette), colorize_mask(o, palette)
d, t, o = d.convert('RGB'), t.convert('RGB'), o.convert('RGB')
[d, t, o] = [self.viz_transform(x) for x in [d, t, o]]
val_img.extend([d, t, o])
val_img = torch.stack(val_img, 0)
val_img = make_grid(val_img.cpu(), nrow=3, padding=5)
if self.config.tensorboard:
self.writer.add_image(f'{self.wrt_mode}/inputs_targets_predictions', val_img, self.wrt_step)
# METRICS TO TENSORBOARD
self.wrt_step = (epoch) * len(self.val_loader)
self.writer.add_scalar(f'{self.wrt_mode}/loss', self.total_loss.average, self.wrt_step)
for k, v in list(seg_metrics.items())[:-1]:
self.writer.add_scalar(f'{self.wrt_mode}/{k}', v, self.wrt_step)
####保存图像
if epoch > 270 and epoch % 50 == 0:
# if epoch % 20 == 0:
ndarr = val_img.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to('cpu', torch.uint8).numpy()
im = Image.fromarray(ndarr)
im.save(os.path.join(self.config.log_dir, str(epoch) + self.config.loss + 'val.png'))
###画confusion
# confusion_matrix(y_true.view(-1).cpu().numpy(), pred.view(-1).cpu().numpy(), labels=range(num_classes),
# normalize='pred')
log = {'loss': self.total_loss.average, **seg_metrics}
return log
def _reset_metrics(self):
self.batch_time = AverageMeter()
self.data_time = AverageMeter()
self.total_loss = AverageMeter()
self.total_inter, self.total_union = 0, 0
self.total_correct, self.total_label = 0, 0
def _update_seg_metrics(self, correct, labeled, inter, union):
self.total_correct += correct
self.total_label += labeled
self.total_inter += inter
self.total_union += union
def _get_seg_metrics(self):
pixAcc = 1.0 * self.total_correct / (np.spacing(1) + self.total_label)
IoU = 1.0 * self.total_inter / (np.spacing(1) + self.total_union)
mIoU = IoU.mean()
return {
"Pixel_Accuracy": np.round(pixAcc, 3),
"Mean_IoU": np.round(mIoU, 3),
"Class_IoU": dict(zip(range(self.num_classes), np.round(IoU, 3)))
}