-
Notifications
You must be signed in to change notification settings - Fork 2
/
classTrainer.py
193 lines (149 loc) · 6.26 KB
/
classTrainer.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
import torch
import torch.nn as nn
from torch.utils import data
from model import PB_FCN, PB_FCN_2
import lr_scheduler
from visualize import LinePlotter
from torchvision.transforms import Compose, Normalize, ToTensor, RandomHorizontalFlip, ColorJitter
from transform import ToYUV, maskLabel
import torchvision.datasets as datasets
import progressbar
import numpy as np
import argparse
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--noScale", help="Use VGA resolution",
action="store_true")
parser.add_argument("--v2", help="Use PB-FCNv2",
action="store_true")
parser.add_argument("--noBall", help="Treat Ball as Background",
action="store_true")
parser.add_argument("--noGoal", help="Treat Goal as Background",
action="store_true")
parser.add_argument("--noRobot", help="Treat Robot as Background",
action="store_true")
parser.add_argument("--noLine", help="Treat Lines as Background",
action="store_true")
args = parser.parse_args()
noScale = args.noScale
v2 = args.v2
nb = args.noBall
ng = args.noGoal
nr = args.noRobot
nl = args.noLine
VGAStr = "VGA" if noScale else ""
v2Str = "v2" if v2 else ""
nbStr = "NoBall" if nb else ""
ngStr = "NoGoal" if ng else ""
nrStr = "NoRobot" if nr else ""
nlStr = "NoLine" if nl else ""
if nb and ng and nr and nl:
print("You need to have at least one non-background class!")
exit(-1)
input_transform = Compose([
ToYUV(),
ToTensor(),
Normalize([.5, 0, 0], [.5, .5, .5]),
])
input_transform_tr = Compose([
RandomHorizontalFlip(),
ColorJitter(brightness=0.5,contrast=0.5,saturation=0.4,hue=0.3),
ToYUV(),
ToTensor(),
Normalize([.5, 0, 0], [.5, .5, .5]),
])
seed = 12345678
np.random.seed(seed)
torch.manual_seed(seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(seed)
batchSize = 64 if v2 else 32
trainloader = data.DataLoader(datasets.ImageFolder("./data/Classification/train/", transform=input_transform_tr),
batch_size=batchSize, shuffle=True, num_workers=6)
valloader = data.DataLoader(datasets.ImageFolder("./data/Classification/val", transform=input_transform),
batch_size=batchSize, shuffle=True,num_workers=6)
numClass = 5 - nb - ng - nr - nl
numFeat = 32
dropout = 0.1
model = PB_FCN_2(True,nClass=numClass) if v2 else PB_FCN(numFeat,numClass,1,noScale,True)
weights = torch.ones(numClass)
if torch.cuda.is_available():
model = model.cuda()
weights = weights.cuda()
criterion = torch.nn.CrossEntropyLoss(weights)
mapLoc = None if torch.cuda.is_available() else {'cuda:0': 'cpu'}
epochs = 200
lr = 1e-2
weight_decay = 1e-5
momentum = 0.9
def cb():
print("Best Model reloaded")
stateDict = torch.load("./pth/bestModel" + VGAStr + v2Str + nbStr + ngStr + nrStr + nlStr + ".pth",
map_location=mapLoc)
model.load_state_dict(stateDict)
optimizer = torch.optim.SGD( [{ 'params': model.parameters()}, ],
lr=lr, momentum=momentum, weight_decay=weight_decay )
scheduler = lr_scheduler.ReduceLROnPlateau(optimizer,'min',factor=0.5,patience=20,verbose=True,threshold=1e-3, cb=cb)
ploter = LinePlotter("RoboCup")
bestLoss = 100
bestAcc = 0
bestTest = 0
for epoch in range(epochs):
model.train()
running_loss = 0.0
running_acc = 0.0
imgCnt = 0
conf = torch.zeros(numClass,numClass).long()
bar = progressbar.ProgressBar(0,len(trainloader),redirect_stdout=False)
for i, (images, labels) in enumerate(trainloader):
if torch.cuda.is_available():
images = images.float().cuda()
labels = labels.cuda()
maskLabel(labels, nb, nr, ng, nl)
optimizer.zero_grad()
pred = torch.squeeze(model(images))
loss = criterion(pred,labels)
loss.backward()
optimizer.step()
bSize = images.size()[0]
imgCnt += bSize
running_loss += loss.item()
_, predClass = torch.max(pred, 1)
running_acc += torch.sum( predClass == labels ).item()*100
for j in range(bSize):
conf[(predClass[j],labels[j])] += 1
bar.update(i)
bar.finish()
print("Epoch [%d] Training Loss: %.4f Training Acc: %.2f" % (epoch+1, running_loss/(i+1), running_acc/(imgCnt)))
#ploter.plot("loss", "train", epoch+1, running_loss/(i+1))
running_loss = 0.0
running_acc = 0.0
imgCnt = 0
conf = torch.zeros(numClass,numClass).long()
model.eval()
bar = progressbar.ProgressBar(0, len(valloader), redirect_stdout=False)
for i, (images, labels) in enumerate(valloader):
if torch.cuda.is_available():
images = images.float().cuda()
labels = labels.cuda()
maskLabel(labels, nb, nr, ng, nl)
pred = torch.squeeze(model(images))
loss = criterion(pred, labels)
bSize = images.size()[0]
imgCnt += bSize
running_loss += loss.item()
_, predClass = torch.max(pred, 1)
running_acc += torch.sum( predClass == labels ).item()*100
for j in range(bSize):
conf[(predClass[j],labels[j])] += 1
bar.update(i)
bar.finish()
print("Epoch [%d] Validation Loss: %.4f Validation Acc: %.2f" % (epoch+1, running_loss/(i+1), running_acc/(imgCnt)))
#ploter.plot("loss", "val", epoch+1, running_loss/(i+1))
if bestLoss > running_loss/(i+1):
bestLoss = running_loss/(i+1)
bestAcc = running_acc/(imgCnt)
print(conf)
torch.save(model.state_dict(), "./pth/bestModel" + VGAStr + v2Str + nbStr + ngStr + nrStr + nlStr + ".pth")
scheduler.step(running_loss/(i+1))
print("Finished: Best Validation Loss: %.4f Best Validation Acc: %.2f" % (bestLoss, bestAcc))