-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
226 lines (188 loc) · 5.97 KB
/
main.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
import os
import sys
import numpy as np
import pandas as pd
from skimage.transform import resize
from skimage.io import imread, imread_collection
from skimage.filters import scharr
from tqdm import tqdm
from random import seed
from unet import train_model, test_model
from config import TRAIN_PATH, TEST_PATH, RANDOM_STATE
from config import IMG_CHAN, IMG_WIDTH, IMG_HEIGHT, MERG_RATION
seed(RANDOM_STATE)
TRAIN_IMAGE_PATTERN = "%s/{}/images/{}.png" % TRAIN_PATH
TRAIN_MASK_PATTERN = "%s/{}/masks/*.png" % TRAIN_PATH
TEST_IMAGE_PATTERN = "%s/{}/images/{}.png" % TEST_PATH
def read_img(img_id, flag_train=True):
if flag_train:
img_path = TRAIN_IMAGE_PATTERN.format(img_id, img_id)
else:
img_path = TEST_IMAGE_PATTERN.format(img_id, img_id)
img = imread(img_path)
img = img[:, :, :3]
#r = scharr(img[:,:,0])
#g = scharr(img[:,:,1])
#b = scharr(img[:,:,2])
#del img
#return np.dstack([r,g,b])
return img
def read_mask(mask_id, shape):
mask_path = TRAIN_MASK_PATTERN.format(mask_id, mask_id)
masks = imread_collection(mask_path).concatenate()
height, width, _ = shape
num_masks = masks.shape[0]
mask = np.zeros((height, width), np.uint32)
dmask = np.zeros((height, width), np.uint32)
for index in range(0, num_masks):
dm = scharr(masks[index])
mask[masks[index] > 0] = 1
dmask[dm > 0] = 1
return np.dstack([mask, dmask])
def read_train_data(ids, cluster):
train_img = []
train_mask = []
for img_id in ids:
ti = read_img(img_id=img_id)
tm = read_mask(mask_id=img_id, shape=ti.shape)
train_img.append(ti)
train_mask.append(tm)
return train_img, train_mask
def read_test_data(ids):
test_img = []
for img_id in ids:
ti = read_img(img_id=img_id, flag_train=False)
test_img.append(ti)
return test_img
def hor_flip(img):
return img[::-1,:,:]
def vert_flip(img):
return img[:,::-1,:]
def pader1(img, shape):
el1 = img[:IMG_WIDTH//2, :IMG_HEIGHT//2, :]
el2 = img[:IMG_WIDTH//2, :shape[1], :]
el1 = np.hstack((el1, el2, el1))
el2 = img[:shape[0], :IMG_HEIGHT//2, :]
print(np.shape(img))
img = np.hstack((el2, img, el2))
print(np.shape(img))
img = np.vstack((el1, img, el1))
return img
def pader2(img, shape):
x,y,c = shape
tx, ty = x//IMG_WIDTH, y//IMG_HEIGHT
img1 = np.hstack((img, img))
img1 = np.vstack((img1, img1))
img = img1[:IMG_WIDTH*(tx+1), :IMG_HEIGHT*(ty+1), :]
return img
def slicer(img):
shape = np.shape(img)
if len(np.shape(img))==2:
img = np.reshape(img, (shape[0], shape[1], 1))
shape = np.shape(img)
img_dict = {
'img': [],
#'vimg': [],
#'himg': [],
#'timg': [],
#'vtimg': [],
#'htimg': [],
'pos': [],
'shape': shape
}
img = pader2(img, shape)
x,y,c = np.shape(img)
tx, ty = x//IMG_WIDTH, y//IMG_HEIGHT
himg = hor_flip(img)
vimg = vert_flip(img)
xflag = False
for i in range(0, MERG_RATION*(tx)):
if i//MERG_RATION >= tx-1 :
sx1 = int(IMG_WIDTH*(i//MERG_RATION))
sx2 = int(IMG_WIDTH*(i//MERG_RATION+1))
xflag = True
else:
sx1 = int(IMG_WIDTH*(i/MERG_RATION))
sx2 = int(IMG_WIDTH*(i/MERG_RATION+1))
yflag = False
for j in range(0, MERG_RATION*(ty)):
if j//MERG_RATION >=ty-1 :
sy1 = int(IMG_HEIGHT*(j//MERG_RATION))
sy2 = int(IMG_HEIGHT*(j//MERG_RATION+1))
yflag = True
else:
sy1 = int(IMG_HEIGHT*(j/MERG_RATION))
sy2 = int(IMG_HEIGHT*(j/MERG_RATION+1))
img_dict['pos'].append([i,j])
#img_dict['vimg'].append(vimg[sx1:sx2, sy1:sy2, :])
#img_dict['himg'].append(himg[sx1:sx2, sy1:sy2, :])
img_dict['img'].append(img[sx1:sx2, sy1:sy2, :])
#sx2, sx1 = x - sx1, x - sx2
#sy2, sy1 = y - sy1, y - sy2
#img_dict['vtimg'].append(vimg[sx1:sx2, sy1:sy2, :])
#img_dict['htimg'].append(himg[sx1:sx2, sy1:sy2, :])
#img_dict['timg'].append(img[sx1:sx2, sy1:sy2, :])
if yflag:
break
if xflag:
break
return img_dict
def img_transform(imgs):
img_set = []
for img in imgs:
img_set.append(slicer(img))
return img_set
def set2list(imgs_set):
imgs = []
for iset in imgs_set:
for l in iset:
if 'img' in l:
imgs.extend(iset[l])
return imgs
def get_train_data(img_ids, cluster):
tr_imgs, masks = read_train_data(img_ids, cluster)
tr_imgs_set = img_transform(tr_imgs)
del tr_imgs
masks_set = img_transform(masks)
del masks
tr_imgs_list = set2list(tr_imgs_set)
del tr_imgs_set
masks_list = set2list(masks_set)
del masks_set
return tr_imgs_list, masks_list
def get_test_data(img_ids):
te_imgs = read_test_data(img_ids)
te_imgs_set = img_transform(te_imgs)
return te_imgs_set
def image_ids_in(root_dir, ignore=[]):
ids = []
for id in os.listdir(root_dir):
if id in ignore:
print('Skipping ID:', id)
else:
ids.append(id)
return ids
def main():
print('Read Path')
#train_data = pd.read_csv('train.csv')
#test_data = pd.read_csv('test.csv')
train_image_ids = image_ids_in('input/train')
#tr_ids = train_data.loc[train_data.hsv_cluster==3, 'image_id'].values
#tr_ids = tr_ids[200:400]
print('Get Data')
X, y = get_train_data(train_image_ids, cluster=3)
print(np.shape(X[0]))
print(np.shape(y[0]))
train_model(X, y, name='3')
del X
del y
del tr_ids
#te_ids = test_data.loc[:, 'image_id'].values
#te_cluster = test_data.loc[:, 'hsv_cluster'].values
#print(len(te_ids))
#test_set = get_test_data(te_ids)
#print(len(test_set))
#test_model(test_set, te_ids, cluster=te_cluster)
#"""
if __name__=='__main__':
main()