-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathsegment.py
190 lines (162 loc) · 6.52 KB
/
segment.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
import json
import numpy as np
import cv2
import os
import tensorflow as tf
from keras import backend as K
from sklearn.model_selection import train_test_split
from keras.models import Model
from keras.regularizers import l2
from keras.layers import Input, Conv2D, MaxPooling2D, UpSampling2D, concatenate ,Conv2DTranspose,BatchNormalization,Dropout
from keras.callbacks import ModelCheckpoint, EarlyStopping,ReduceLROnPlateau
from keras.losses import binary_crossentropy
from keras.optimizers import Adam
from keras.metrics import MeanIoU
from keras.backend import epsilon
from keras.preprocessing.image import ImageDataGenerator
from keras.models import save_model
def dice_loss(y_true, y_pred, epsilon=1e-6):
intersection = tf.reduce_sum(y_true * y_pred)
union = tf.reduce_sum(y_true + y_pred)
dice_loss = 1 - (2 * intersection + epsilon) / (union + epsilon)
return dice_loss
def dice_coef(y_true, y_pred):
y_true_f = K.flatten(y_true)
y_pred_f = K.flatten(y_pred)
intersection = K.sum(y_true_f * y_pred_f)
return (2.0 * intersection + K.epsilon()) / (K.sum(y_true_f) + K.sum(y_pred_f) + K.epsilon())
def unet(input_shape, dropout_rate=0.2, l2_regularization=1e-4):
# Encoder
inputs = Input(input_shape)
conv1 = Conv2D(32, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(inputs)
conv1 = BatchNormalization()(conv1)
pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
conv2 = Conv2D(64, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(pool1)
conv2 = BatchNormalization()(conv2)
pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
conv3 = Conv2D(128, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(pool2)
conv3 = BatchNormalization()(conv3)
pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
conv4 = Conv2D(256, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(pool3)
conv4 = BatchNormalization()(conv4)
pool4 = MaxPooling2D(pool_size=(2, 2))(conv4)
# Decoder
up5 = Conv2DTranspose(128, 2, strides=(2, 2), padding='same')(conv4)
merge5 = concatenate([conv3, up5], axis=3)
conv5 = Conv2D(128, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(merge5)
conv5 = BatchNormalization()(conv5)
conv5 = Dropout(dropout_rate)(conv5)
up6 = Conv2DTranspose(64, 2, strides=(2, 2), padding='same')(conv5)
merge6 = concatenate([conv2, up6], axis=3)
conv6 = Conv2D(64, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(merge6)
conv6 = BatchNormalization()(conv6)
conv6 = Dropout(dropout_rate)(conv6)
up7 = Conv2DTranspose(32, 2, strides=(2, 2), padding='same')(conv6)
merge7 = concatenate([conv1, up7], axis=3)
conv7 = Conv2D(32, 3, activation='relu', padding='same', kernel_regularizer=l2(l2_regularization))(merge7)
conv7 = BatchNormalization()(conv7)
conv7 = Dropout(dropout_rate)(conv7)
outputs = Conv2D(1, 1, activation='sigmoid')(conv7)
model = Model(inputs=inputs, outputs=outputs)
return model
batch_size = 2
# Define directories for images and masks
image_dir = 'Unet Data/images'
mask_dir = 'Unet Data/masks'
# print(os.path.exists(image_dir))
# print(os.listdir(image_dir))
#
# print(os.path.exists(mask_dir))
# print(os.listdir(mask_dir))
#
# train_imggen = ImageDataGenerator(
# rescale=1./255,
# validation_split=0.2
# # Specify the validation split
# )
# train_maskgen = ImageDataGenerator(
# rescale=1./255,
# validation_split=0.2
#
# # Specify the validation split
# )
# # Create a data generator for the training set
# train_generator = train_imggen.flow_from_directory(
# directory=image_dir,
# target_size=(224, 224),
# batch_size=batch_size,
# class_mode='input',
# subset='training',
# color_mode='grayscale',
# seed=1# Specify the subset as training set
# )
#
# # Create a data generator for the validation set
# val_generator = train_imggen.flow_from_directory(
# directory= image_dir,
# target_size=(224, 224),
# batch_size=batch_size,
# class_mode='input',
# subset='validation',
# color_mode='rgb',
# seed=1# Specify the subset as validation set
# )
#
# train_maskgenerator = train_maskgen.flow_from_directory(
# directory=mask_dir,
# target_size=(224, 224),
# batch_size=batch_size,
# class_mode=None,
# subset='training',
# color_mode='rgb',
# seed=1# Specify the subset as training set
# )
#
# # Create a data generator for the validation set
# val_maskgenerator = train_maskgen.flow_from_directory(
# directory=mask_dir,
# target_size=(224, 224),
# batch_size=batch_size,
# class_mode=None,
# subset='validation',
# seed=1 # Specify the subset as validation set
# )
#
# train_dataset = zip(train_generator, train_maskgenerator)
# val_dataset = zip(val_generator, val_maskgenerator)
# Get list of file names in image and mask directories
image_files = os.listdir( image_dir )
mask_files = os.listdir( mask_dir )
# Create arrays to hold images and masks
images = []
masks = []
# Iterate over files and load images and masks
for image_file, mask_file in zip(image_files, mask_files):
# Load image
image = cv2.imread(os.path.join(image_dir, image_file), cv2.IMREAD_GRAYSCALE)
image = cv2.resize(image, (224, 224))
image = np.expand_dims(image, axis=-1)
images.append(image)
# Load mask
mask = cv2.imread(os.path.join(mask_dir, mask_file), cv2.IMREAD_GRAYSCALE)
mask = cv2.resize(mask, (224, 224))
mask = np.expand_dims(mask, axis=-1)
masks.append(mask)
# Convert data to numpy arrays
images = np.array( images )
masks = np.array( masks )
# Normalize images to [0, 1]
images = images / 255.0
masks = masks/255.0
# Split data into training and validation sets
train_images, val_images, train_masks, val_masks = train_test_split(images, masks, test_size=0.2)
callbacks = [
EarlyStopping(monitor='val_loss', min_delta=0.001, patience=5, mode='min', restore_best_weights=True),
ModelCheckpoint(filepath='unet_weights.h5', monitor='val_loss', save_best_only=True, mode='min'),
ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=5, min_lr=0.001)
]
optimizer = Adam(learning_rate=0.001,beta_1 = 0.9,beta_2 = 0.999,epsilon = 1e-7)
model = unet(input_shape=(224, 224, 1))
model.compile(optimizer=optimizer, loss=dice_loss, metrics=[dice_coef,'accuracy'])
history = model.fit(train_images,train_masks, validation_data=(val_images,val_masks), epochs=75, batch_size=batch_size,callbacks=callbacks)
model.save('segment.h5')