-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
241 lines (223 loc) · 12.1 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
from config import config
import torch
import torch.nn as nn
from model.model_one import *
from model.unet import *
from model.pspnet import *
from model.segnet import *
from model.bisenet import *
from model.deeplab import *
from model.loss import *
from tqdm import tqdm, trange
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.utils.data import DataLoader, SequentialSampler, DistributedSampler
from torch.cuda.amp import autocast, GradScaler
from data.dataset import loadedDataset
import cv2
import time
import random
def main(args):
# ---------------------------set GPU environment---------------------------
n_gpu = torch.cuda.device_count()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
local_rank = args.local_rank
torch.cuda.set_device(local_rank)
torch.distributed.init_process_group('nccl')
seed = 42
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
if n_gpu > 0:
torch.cuda.manual_seed_all(seed)
# ---------------------------dataset---------------------------
# train
train_dataset = loadedDataset(txt = "./datasets/train/train_example.txt")
train_sampler = DistributedSampler(train_dataset)
train_loader = DataLoader(train_dataset, batch_size = args.train_batch_size // n_gpu, sampler = train_sampler,
num_workers = args.num_workers, drop_last = False, pin_memory = True)
# validation
val_dataset = loadedDataset(txt = "./datasets/val/val_example.txt")
val_sampler = SequentialSampler(val_dataset)
val_loader = DataLoader(val_dataset, sampler = val_sampler, batch_size = args.val_batch_size,
num_workers = args.num_workers, drop_last = True, pin_memory = True)
# test
test_dataset = loadedDataset(txt = "./datasets/test/test_example.txt")
test_sampler = SequentialSampler(test_dataset)
test_loader = DataLoader(test_dataset, sampler = test_sampler, batch_size = args.test_batch_size,
num_workers = args.num_workers, drop_last = True, pin_memory = True)
# ---------------------------model---------------------------
if args.model_name == "unet":
model = unet(3,2)
elif args.model_name == "fcn":
model = fcn()
elif args.model_name == "pspnet":
model = PSPNet()
elif args.model_name == "segnet":
model = segnet()
elif args.model_name == "deeplab":
model = deeplab()
# =============train=============== #
if args.mode == "train":
if args.model_name == "model_one":
model = model_one(n_classes=2, aux_mode=args.mode)
elif args.model_name == "bisenet":
model = bisenet(n_classes=2, aux_mode='eval')
model.to(device)
model = DDP(model, find_unused_parameters=True,
device_ids=[local_rank], output_device=local_rank)
if args.load_model:
pretrained_dict = torch.load(args.pretrained_model)
model_dict = model.state_dict()
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
model_dict.update(pretrained_dict)
model.load_state_dict(model_dict)
# calculate the number of parameters
if local_rank == 0:
total_params = sum(p.numel() for p in model.parameters())
print(f'{total_params:,} total parameters.')
total_trainable_params = sum(
p.numel() for p in model.parameters() if p.requires_grad)
print(f'{total_trainable_params:,} training parameters.')
# scaler = GradScaler()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr, weight_decay=1e-4)
#optimizer = torch.optim.SGD(model.parameters(), lr=args.lr, weight_decay=0.0001, momentum=0.9)
StepLR = torch.optim.lr_scheduler.LambdaLR(optimizer,lr_lambda=lambda epoch: (1-epoch/args.num_train_epochs)**0.9)
criterion1 = simpleloss()
criterion2 = BinaryDiceLoss()
crtierion_edge = DetailAggregateLoss()
criterion_semantic_and_result = WeightedOhemCELoss(thresh=0.7, n_min=256*256*args.train_batch_size//(16*n_gpu), num_classes=2)
g = lambda x, y: criterion1(x,y)
g1 = lambda x, y: crtierion_edge(x,y)
g2 = lambda x, y: criterion_semantic_and_result(x,y)
f = open("./" + args.model_name + "_train_result.txt","w+")
for epoch_ in trange(args.num_train_epochs, desc="Epoch"):
for step, (img, lab) in enumerate(tqdm(train_loader, desc="Iteration", ascii=True, total = len(train_loader))):
# train
data, target = img.type(torch.cuda.FloatTensor).to(device), lab.type(torch.cuda.FloatTensor).to(device)
_, _, h, w = data.shape
a = h % 32 // 2
b = w % 32 // 2
data, target = data[:, :, a : h - a, b : w - b], target[:, a : h - a, b : w - b]
if local_rank == 0:
start_time = time.time()
#with autocast():
predict = model(data)
#predict = torch.argmax(predict, dim=1)
if args.model_name == "model_one":
detail, semantic, predict = model(data)
loss = 0.3 * g1(detail, target) + 0.8 * (g(semantic, target) + g(predict, target)) + 0.2 * (g2(semantic, target) + g2(predict, target))
elif args.model_name == "bisenet":
predict = model(data)
loss = g(predict, target)
else:
loss = g(predict, target)
if n_gpu > 1:
loss = loss.mean()
optimizer.zero_grad()
loss.backward()
optimizer.step()
if local_rank == 0:
delta_time = time.time() - start_time
if (local_rank == 0) and ((step + 1) % args.eval_steps == 0):
print("===================train===================")
predict2 = torch.argmax(predict, dim=1)
acc = predict2.eq(target.view_as(predict2)).sum().item() / target.numel()
TP = predict2[target > 0.5].eq(target[target > 0.5].view_as(predict2[target > 0.5])).sum().item()
FN = target[target > 0.5].numel() - TP
TN = predict2[target < 0.5].eq(target[target < 0.5].view_as(predict2[target < 0.5])).sum().item()
FP = target[target < 0.5].numel() - TN
print(TP)
print(FN)
print(TN)
print(FP)
precision = TP / (TP + FP + 1e-6)
recall = TP / (TP + FN + 1e-6)
IOU = TP / (TP + FN + FP + 1e-6)
F1_score = 2 * precision * recall / (precision + recall + 1e-6)
print("loss={:.6f}, acc={:.6f}, precision={:.6f}, recall={:.6f}, F1_score={:.6f}, iou={:.6f}".format(
loss, acc, precision, recall, F1_score, IOU))
f.write("Epoch = " + str(epoch_) + ", step = " + str(step) + ", acc = " + str(acc) +
", precision = " + str(precision) + ", recall = " + str(recall) + ", loss = " + str(loss) +
", F1_score = " + str(F1_score) + ", time = " + str(delta_time) + ", IOU = " + str(IOU)
+ ", TP = " + str(TP) + ", FN = " + str(FN) + ", TN" + str(TN) + ", FP = " + str(FP) + "\n")
# eval
if (step + 1) % args.eval_steps == 0:
print("===================Evaluate===================")
model.eval()
for k, (img1, lab1) in enumerate(tqdm(val_loader, desc="Evaluating", total = len(val_loader))):
data = img1.type(torch.cuda.FloatTensor).to(device)
target = lab1.type(torch.cuda.FloatTensor).to(device)
_, _, h, w = data.shape
a = h % 32 // 2
b = w % 32 // 2
data, target = data[:, :, a: h - a, b: w - b], target[:, a: h - a, b: w - b]
with torch.no_grad():
if args.model_name == "model_one":
_, _, predict = model(data)
#loss = args.deep_supervision_weight*(g(sub1, target)+g(sub2, target)+g(sub3, target)) \
# + g(predict, target)
elif args.model_name == "bisenet":
#predict, _, _, _, _ = model(data)
predict = model(data)
else:
predict = model(data)
loss = g(predict, target)
predict = torch.argmax(predict, dim=1)
for i in range(args.val_batch_size):
img1 = target[i,:,:].cpu().detach().numpy()*255
file1 = "./example/" + \
"_step" + str(k) + "_item" + str(i) + "label.jpg"
cv2.imwrite(file1, img1)
img2 = predict[i,:,:].cpu().detach().numpy()*255
file2 = "./example/" + \
"_step" + str(k) + "_item" + str(i) + "predict.jpg"
cv2.imwrite(file2, img2)
img3 = data[i,:,:,:].permute(1,2,0).cpu().detach().numpy()*255
file3 = "./example/" + \
"_step" + str(k) + "_item" + str(i) + "original_data.jpg"
cv2.imwrite(file3, img3)
StepLR.step()
args.deep_supervision_weight = args.deep_supervision_weight * (1 - (epoch_ + 1) / args.num_train_epochs)
if local_rank == 0:
PATH = "./model/" + args.model_name + "_epoch" + str(epoch_) + ".pth"
torch.save(model.state_dict(), PATH)
f.close()
elif args.mode == "test":
if args.model_name == "model_one":
model = model_one(n_classes=2, aux_mode=args.mode)
elif args.model_name == "bisenet":
model = bisenet(n_classes=2, aux_mode='eval')
model.to(device)
model = DDP(model, find_unused_parameters=True,
device_ids=[local_rank], output_device=local_rank)
#model.load_state_dict(torch.load(args.pretrained_model), strict=False)
pretrained_dict = torch.load(args.pretrained_model)
model_dict = model.state_dict()
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
model_dict.update(pretrained_dict)
model.load_state_dict(model_dict)
model.eval()
f = open("./" + args.model_name + "_test_result.txt", "w+")
begin = time.time()
acc, precision, recall, F1_score, IOU, delta_time = 0, 0, 0, 0, 0,0
total = 0
for step, (img, lab) in enumerate(tqdm(test_loader, desc="Iteration", ascii=True, total = len(test_loader))):
data, target = img.type(torch.cuda.FloatTensor).to(device), lab.type(torch.cuda.FloatTensor).to(device)
_, _, h, w = data.shape
a = h % 32
b = w % 32
data = data[:, :, :h-a, :w-b]
if local_rank == 0:
start_time = time.time()
with torch.no_grad():
predict = model(data)
delta_time += time.time() - begin
predict = torch.argmax(predict, dim=1)
img2 = predict.squeeze().cpu().detach().numpy() * 255
file2 = "./datasets/demo/" + str(total) + "predict.jpg"
cv2.imwrite(file2, img2)
total += 1
f.close()
if __name__ == '__main__':
args = config()
main(args)