-
Notifications
You must be signed in to change notification settings - Fork 0
/
ISIC2018_dataset.py
108 lines (90 loc) · 4.85 KB
/
ISIC2018_dataset.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
import os
import fnmatch
import numpy as np
import random
import cv2 as cv
from keras.utils.np_utils import to_categorical
import pandas as pd
np.random.seed(4)
import pdb
mean_imagenet = [123.68, 103.939, 116.779] # rgb
mean_train_isic2017 = np.array([[[ 180.71656799]],[[ 151.13494873]],[[ 139.89967346]]]);
std_train_isic2017 = train_std = np.array([[[1]],[[1]],[[ 1]]]);
def get_labels(image_list, csv_file):
image_list = [filename.split('.')[0] for filename in image_list]
return to_categorical(pd.read_csv(csv_file,index_col=0).loc[image_list]['label'].values.flatten())
def split_isic_train(train_folder, split_ratio):
imglist = fnmatch.filter(os.listdir(train_folder), '*.jpg')
len_imglist = len(imglist)
num_each = len_imglist/sum(split_ratio)
index = list(range(len_imglist))
#random.shuffle(index)
trainlist = index[0: split_ratio[0] * num_each]
validationlist = index[split_ratio[0] * num_each: (split_ratio[0] + split_ratio[1])*num_each]
testlist = index[(split_ratio[0] + split_ratio[1]) * num_each :]
train_list = [imglist[i] for i in trainlist]
validation_list = [imglist[i] for i in validationlist]
test_list = [imglist[i] for i in testlist]
return train_list, validation_list, test_list
def resize_image(ori_folder, ori_mask_folder, ori_list, resize_image_folder, resize_mask_folder, height, width):
#pdb.set_trace()
os.makedirs(resize_image_folder)
os.makedirs(resize_mask_folder)
for imgname in ori_list:
ori_image = cv.imread(os.path.join(ori_folder, imgname))
ori_mask = cv.imread(os.path.join(ori_mask_folder, imgname.replace(".jpg", "_segmentation.png")), cv.IMREAD_GRAYSCALE)
_, ori_mask = cv.threshold(ori_mask,127,255,cv.THRESH_BINARY)
resize_image = cv.resize(ori_image, (height, width), interpolation=cv.INTER_CUBIC)
resize_mask = cv.resize(ori_mask, (height, width), interpolation=cv.INTER_CUBIC)
_, resize_mask = cv.threshold(resize_mask,127,255,cv.THRESH_BINARY)
cv.imwrite(os.path.join(resize_image_folder, imgname.replace(".jpg", ".png")), resize_image)
cv.imwrite(os.path.join(resize_mask_folder, imgname.replace(".jpg", ".png")), resize_mask)
def resize_only_image(ori_folder, ori_list, resize_image_folder, height, width):
#pdb.set_trace()
os.makedirs(resize_image_folder)
shape_dict = {}
for imgname in ori_list:
ori_image = cv.imread(os.path.join(ori_folder, imgname))
shape_dict[imgname] = (ori_image.shape[1], ori_image.shape[0])
resize_image = cv.resize(ori_image, (height, width), interpolation=cv.INTER_CUBIC)
cv.imwrite(os.path.join(resize_image_folder, imgname.replace(".jpg", ".png")), resize_image)
return shape_dict
def load_image(image_folder, mask_folder, image_list, height, width, remove_mean_imagenet, remove_mean_dataset, rescale_mask):
n_channel = 3
img_array = np.zeros((len(image_list), n_channel, height, width), dtype=np.float32)
img_mask_array = np.zeros((len(image_list), height, width), dtype=np.float32)
for i in range(len(image_list)):
image_path = os.path.join(image_folder, image_list[i].replace(".jpg",".png"))
mask_path = os.path.join(mask_folder, image_list[i].replace(".jpg",".png"))
img = cv.imread(image_path)
if not os.path.exists(mask_path):
print(image_path)
continue
img = cv.cvtColor(img, cv.COLOR_BGR2RGB).astype(np.float32)
if remove_mean_imagenet:
for channel in [0, 1, 2]:
img[:,:,channel] -= mean_imagenet[channel]
img = img.transpose((2,0,1)).astype(np.float32)
if remove_mean_dataset:
img = (img - mean_train_isic2017)/std_train_isic2017
img_array[i] = img
img_mask = cv.imread(mask_path, cv.IMREAD_GRAYSCALE)
_, img_mask_array[i] = cv.threshold(img_mask,127,255,cv.THRESH_BINARY)
if rescale_mask:
img_mask_array[i] = img_mask_array[i]/255.
return (img_array, img_mask_array.astype(np.uint8).reshape((img_mask_array.shape[0],1,img_mask_array.shape[1],img_mask_array.shape[2])))
def load_only_image(image_folder, image_list, height, width, remove_mean_imagenet, remove_mean_dataset):
n_channel = 3
img_array = np.zeros((len(image_list), n_channel, height, width), dtype=np.float32)
for i in range(len(image_list)):
image_path = os.path.join(image_folder, image_list[i].replace(".jpg",".png"))
img = cv.imread(image_path)
img = cv.cvtColor(img, cv.COLOR_BGR2RGB).astype(np.float32)
if remove_mean_imagenet:
for channel in [0, 1, 2]:
img[:,:,channel] -= mean_imagenet[channel]
img = img.transpose((2,0,1)).astype(np.float32)
if remove_mean_dataset:
img = (img - mean_train_isic2017)/std_train_isic2017
img_array[i] = img
return img_array