-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathutils.py
181 lines (150 loc) · 6.29 KB
/
utils.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
import random
import numpy as np
import torch
import pandas as pd
import os
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset
import cv2
import albumentations as A
from albumentations.pytorch import ToTensorV2
from sklearn.model_selection import StratifiedKFold, StratifiedGroupKFold
from pathlib import Path
from glob import glob
from torch.utils.data import DataLoader
from glob import glob
class SkinDataSet(Dataset):
def __init__(self, df, cat, transforms=None):
self.cat = cat
if cat == 'train':
self.df_positive = df[df['target'] == 1].reset_index()
self.df_negative = df[df['target'] == 0].reset_index()
self.file_names_positive = self.df_positive['file_path'].values
self.file_names_negative = self.df_negative['file_path'].values
self.targets_positive = self.df_positive['target'].values
self.targets_negative = self.df_negative['target'].values
self.transforms = transforms
else:
self.df = df
self.file_names = df['file_path'].values
self.targets = df['target'].values
self.transforms = transforms
def __len__(self):
if self.cat == 'train':
return len(self.df_positive) * 2
else:
return len(self.df)
def __getitem__(self, index):
if self.cat == 'train':
if random.random() >= 0.5:
df = self.df_positive
file_names = self.file_names_positive
targets = self.targets_positive
else:
df = self.df_negative
file_names = self.file_names_negative
targets = self.targets_negative
index = index % df.shape[0]
img_path = file_names[index]
target = targets[index]
else:
img_path = self.file_names[index]
target = self.targets[index]
img = cv2.imread(img_path)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
if self.transforms:
img = self.transforms(image=img)['image']
return {'image': img, 'target': target}
def set_seed(seed):
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
# When running on the CuDNN backend, two further options must be set
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# Set a fixed value for the hash seed
os.environ['PYTHONHASHSEED'] = str(seed)
def criterion(outputs, targets):
return nn.CrossEntropyLoss()(outputs, targets)
def prepare_loaders(df, train_batch_size, val_batch_size, img_size, fold=0):
df_train = df[df.kfold != fold].reset_index(drop=True)
df_valid = df[df.kfold == fold].reset_index(drop=True)
data_transforms = {
"train": A.Compose([
A.Transpose(p=0.5),
A.VerticalFlip(p=0.5),
A.HorizontalFlip(p=0.5),
A.RandomBrightnessContrast(limit=0.2, p=0.75),
A.OneOf([
A.MotionBlur(blur_limit=5),
A.MedianBlur(blur_limit=5),
A.GaussianBlur(blur_limit=5),
A.GaussNoise(var_limit=(5.0, 30.0)),
], p=0.7),
A.CLAHE(clip_limit=4.0, p=0.7),
A.Resize(img_size, img_size),
A.RandomRotate90(p=0.5),
A.Flip(p=0.5),
A.Downscale(p=0.25),
A.ShiftScaleRotate(shift_limit=0.1,
scale_limit=0.15,
rotate_limit=15,
border_mode=0,
p=0.85),
A.HueSaturationValue(
hue_shift_limit=10,
sat_shift_limit=20,
val_shift_limit=10,
p=0.5
),
A.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225],
max_pixel_value=255.0,
p=1.0
),
ToTensorV2()], p=1.),
"valid": A.Compose([
A.Resize(img_size, img_size),
A.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225],
max_pixel_value=255.0,
p=1.0
),
ToTensorV2()], p=1.)
}
train_dataset = SkinDataSet(df_train, cat='train', transforms=data_transforms['train'])
valid_dataset = SkinDataSet(df_valid, cat='valid', transforms=data_transforms['valid'])
train_loader = DataLoader(train_dataset, batch_size=train_batch_size,
shuffle=True, pin_memory=True, drop_last=True, num_workers=4)
valid_loader = DataLoader(valid_dataset, batch_size=val_batch_size,
shuffle=False, pin_memory=True, num_workers=4)
return train_loader, valid_loader
def df_preprocess(nfold, fold):
path = Path('../data')
train_img_path = path /'train-image/image'
train_images = glob(str(train_img_path) + '/*')
train = pd.read_csv(path / 'train-metadata.csv', low_memory=False)
train_df1 = pd.read_csv(path / 'train_2019.csv', low_memory=False)
train_df2 = pd.read_csv(path / 'train_mealanoma.csv', low_memory=False)
# Count the Images
def get_img_path(image_id):
return f"{train_img_path}/{image_id}.jpg"
train_df1.rename(columns={'image_name':'isic_id'}, inplace=True)
train_df2.rename(columns={'image_name':'isic_id'}, inplace=True)
train_df1_positive = train_df1[train_df1["target"] == 1].reset_index(drop=True)
train_df2_positive = train_df2[train_df2['target'] == 1].reset_index(drop=True)
train_positive = train[train['target'] == 1].reset_index(drop=True)
train_negative = train[train["target"] == 0].reset_index(drop=True)
train_positive_df = pd.concat([train_df1_positive, train_df2_positive, train_positive],axis=0, ignore_index=True)
train_positive_df.isna().sum()
df = pd.concat([train_positive_df, train_negative.iloc[:train_positive_df.shape[0] :]])
df['file_path'] = df['isic_id'].apply(get_img_path)
df = df[ df["file_path"].isin(train_images) ].reset_index(drop=True)
sgkf = StratifiedGroupKFold(n_splits=nfold)
for fold, ( _, val_) in enumerate(sgkf.split(df, df.target,df.patient_id)):
df.loc[val_ , "kfold"] = int(fold)
return df