-
Notifications
You must be signed in to change notification settings - Fork 4
/
train.py
156 lines (137 loc) · 5.78 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
# -*- coding: utf-8 -*-
"""
Created on Thu Dec 21 13:14:50 2017
@author: B
"""
import sys
sys.path.append('/root/tls/Models/')
import numpy as np
np.random.seed(1006)
import argparse
import Models , PageLoadBatches
from keras.callbacks import ModelCheckpoint
from keras import optimizers
import cv2
import os
import random
fold=3
parser = argparse.ArgumentParser()
parser.add_argument("--n_classes", type=int, default = 2 )
parser.add_argument("--input_height", type=int , default = 320 )
parser.add_argument("--input_width", type=int , default = 320 )
parser.add_argument("--epochs", type = int, default = 250 )
parser.add_argument("--batch_size", type = int, default = 16 )
parser.add_argument("--model_name", type = str , default = "fcn8")
parser.add_argument("--optimizer_name", type = str , default = "sgd")
parser.add_argument("--load_weights", type = str , default = '')
args = parser.parse_args()
train_batch_size = args.batch_size
val_batch_size = args.batch_size
n_classes = args.n_classes
input_height = args.input_height
input_width = args.input_width
epochs = args.epochs
load_weights = args.load_weights
optimizer_name = args.optimizer_name
model_name = args.model_name
trains=[list(range(0,20)),
list(range(5,25)),
list(range(10,30)),
[15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,0,1,2,3,4],
[20,21,22,23,24,25,26,27,28,29,0,1,2,3,4,5,6,7,8,9],
[25,26,27,28,29,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14]]
validations=[list(range(20,25)),
list(range(25,30)),
list(range(0,5)),
list(range(5,10)),
list(range(10,15)),
list(range(15,20))]
tests=[list(range(25,30)),
list(range(0,5)),
list(range(5,10)),
list(range(10,15)),
list(range(15,20)),
list(range(20,25))]
print('training for fold '+str(fold))
if not (os.path.exists('ptrain'+str(fold))):
print('create patch folders if does not exist')
os.mkdir('ptrain'+str(fold))
os.mkdir('pltrain'+str(fold))
os.mkdir('pvalidation'+str(fold))
os.mkdir('plvalidation'+str(fold))
print('train and validation patch folders are generated')
print('train patches are being generated')
patchSize=320
pages=[]
labels=[]
for page in trains[fold]:
pages.append('data/pages/'+str(page)+'.png')
for label in trains[fold]:
labels.append('data/labels/'+str(label)+'.png')
print('train pages are: ')
print(pages)
print('train labels are: ')
print(labels)
i=0
while (i <50000):
page_number=random.randint(0,19)
page_name=pages[page_number]
label_name=labels[page_number]
page=cv2.imread(page_name,0)
lpage=cv2.imread(label_name,0)
rows,cols=page.shape
x=random.randint(0,rows-patchSize)
y=random.randint(0,cols-patchSize)
patch=page[x:x+patchSize,y:y+patchSize]
cv2.imwrite('ptrain'+str(fold)+'/'+page_name.split('/')[2][:-4]+"_patch"+str(i)+".png",patch)
lpatch=lpage[x:x+patchSize,y:y+patchSize]
cv2.imwrite('pltrain'+str(fold)+'/'+label_name.split('/')[2][:-4]+"_patch"+str(i)+".png",lpatch)
i=i+1
print(str(i)+' train patches for fold '+str(fold)+ ' is generated')
print('validation patches are being generated')
pages=[]
labels=[]
for page in validations[fold]:
pages.append('data/pages/'+str(page)+'.png')
for label in validations[fold]:
labels.append('data/labels/'+str(label)+'.png')
print('validation pages are: ')
print(pages)
print('validation labels are: ')
print(labels)
i=0
while (i <6000):
page_number=random.randint(0,4)
page_name=pages[page_number]
label_name=labels[page_number]
page=cv2.imread(page_name,0)
lpage=cv2.imread(label_name,0)
rows,cols=page.shape
x=random.randint(0,rows-patchSize)
y=random.randint(0,cols-patchSize)
patch=page[x:x+patchSize,y:y+patchSize]
cv2.imwrite('pvalidation'+str(fold)+'/'+page_name.split('/')[2][:-4]+"_patch"+str(i)+".png",patch)
lpatch=lpage[x:x+patchSize,y:y+patchSize]
cv2.imwrite('plvalidation'+str(fold)+'/'+label_name.split('/')[2][:-4]+"_patch"+str(i)+".png",lpatch)
i=i+1
print(str(i)+' validation patches for fold '+str(fold)+ ' is generated')
modelFns = { 'vgg_segnet':Models.VGGSegnet.VGGSegnet , 'vgg_unet':Models.VGGUnet.VGGUnet , 'vgg_unet2':Models.VGGUnet.VGGUnet2 , 'fcn8':Models.FCN8.FCN8 , 'fcn32':Models.FCN32.FCN32 }
modelFN = modelFns[ model_name ]
m = modelFN( n_classes , input_height=input_height, input_width=input_width )
sgd = optimizers.SGD(lr=0.001)
#adm=optimizers.Adam(lr=0.0001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=5e-05)
m.compile(loss='categorical_crossentropy',
optimizer= sgd,
metrics=['accuracy'])
if len( load_weights ) > 0:
print("loading initial weights")
m.load_weights(load_weights)
print ( m.output_shape)
output_height = m.outputHeight
output_width = m.outputWidth
G = PageLoadBatches.imageSegmentationGenerator( 'ptrain'+str(fold)+'/', 'pltrain'+str(fold)+'/' , train_batch_size, n_classes , input_height , input_width , output_height , output_width )
G2 = PageLoadBatches.imageSegmentationGenerator( 'pvalidation'+str(fold)+'/' , 'plvalidation'+str(fold)+'/' , val_batch_size, n_classes , input_height , input_width , output_height , output_width )
mcp=ModelCheckpoint( filepath='bestweights'+str(fold), monitor='val_loss', save_best_only=True, save_weights_only=True,verbose=1)
for ep in range( epochs ):
print ('epoch:'+str(ep))
m.fit_generator( G , 3125, validation_data=G2 , validation_steps=375, epochs=1,callbacks=[mcp] )