-
Notifications
You must be signed in to change notification settings - Fork 0
/
1.2.UnetBinary.py
162 lines (134 loc) · 4.58 KB
/
1.2.UnetBinary.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
from torch import nn
from torch.nn import functional as F
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
import numpy as np
from utils.dataset import *
from utils.callbacks import DiceCallback as MyDiceCallbak, IouCallback as MyIouCallback, AccuracyCallback
from catalyst.dl import SupervisedRunner, DiceCallback, IouCallback, AUCCallback
from catalyst.utils import set_global_seed, prepare_cudnn
import segmentation_models_pytorch as smp
from utils.mobilenetv3 import mobilenetv3
from tqdm import tqdm
import os
prepare_cudnn(True, True)
set_global_seed(0)
NAME = '1.2.resnet50_binary_hard_transforms'
LOGDIR = f"./logdir/{NAME}"
# Train binary
# ------------------------------------------------------------------------------
logdir = os.path.join(LOGDIR, 'binary/')
num_epochs = 40
FP16 = True
batch_size = 24
default_batch_size = 8
lr = 1e-4 * batch_size / default_batch_size
weight_decay = 1e-5
momentum = 0.9
mean = (0.485, 0.456, 0.406)
std = (0.229, 0.224, 0.225)
num_workers = 6
model = mobilenetv3(1)
# Dataloaders
train, val = get_train_val_dataloaders(df='dataset/train.csv',
data_folder='dataset/train_images',
mean=mean,
std=std,
batch_size=batch_size,
num_workers=num_workers,
pin_memory=False,
binary=True,
hard_transforms=True)
loaders = {"train": train, "valid": val}
# Optimizer
criterion = nn.BCEWithLogitsLoss()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
scheduler = ReduceLROnPlateau(optimizer, mode="min", patience=3, verbose=True)
# Train
runner = SupervisedRunner()
runner.train(
model=model,
criterion=criterion,
optimizer=optimizer,
scheduler=scheduler,
loaders=loaders,
logdir=logdir,
num_epochs=num_epochs,
verbose=True,
callbacks=[
AUCCallback(),
AccuracyCallback(threshold=0.5),
],
fp16=FP16,
)
del model, train, val, criterion, optimizer, scheduler, runner
# ------------------------------------------------------------------------------
# Train segmentation
# ------------------------------------------------------------------------------
num_epochs = 50
encoder = 'resnet50'
logdir = os.path.join(LOGDIR, 'seg/')
FP16 = True
batch_size = 8
default_batch_size = 8
lr = 1e-4 * batch_size / default_batch_size
weight_decay = 1e-5
momentum = 0.9
# Dataloaders
train, val = get_train_val_dataloaders(df='dataset/train.csv',
data_folder='dataset/train_images',
mean=mean,
std=std,
batch_size=batch_size,
num_workers=num_workers,
pin_memory=False,
binary=False,
hard_transforms=True,
only_has_mask=True)
loaders = {"train": train, "valid": val}
# Model
model = smp.Unet(encoder, encoder_weights='imagenet', classes=4, activation=None)
# Optimizer
# criterion = nn.BCEWithLogitsLoss()
criterion = nn.BCEWithLogitsLoss()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
scheduler = ReduceLROnPlateau(optimizer, mode="min", patience=3, verbose=True)
# Train
runner = SupervisedRunner()
runner.train(
model=model,
criterion=criterion,
optimizer=optimizer,
scheduler=scheduler,
loaders=loaders,
logdir=logdir,
num_epochs=num_epochs,
verbose=True,
callbacks=[
DiceCallback(threshold=0.5, prefix='catalyst_dice'),
IouCallback(threshold=0.5, prefix='catalyst_iou'),
MyDiceCallbak(threshold=0.5),
MyIouCallback(threshold=0.5),
],
fp16=FP16,
)
del train, optimizer
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
runner.train(
model=model,
criterion=criterion,
optimizer=optimizer,
scheduler=scheduler,
loaders={"train": val, "valid": val},
logdir=logdir,
num_epochs=num_epochs,
verbose=True,
callbacks=[
DiceCallback(threshold=0.5, prefix='catalyst_dice'),
IouCallback(threshold=0.5, prefix='catalyst_iou'),
MyDiceCallbak(threshold=0.5),
MyIouCallback(threshold=0.5),
],
fp16=FP16,
)
# ------------------------------------------------------------------------------