-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataloder.py
254 lines (210 loc) · 9.75 KB
/
dataloder.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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
import os
import tensorflow as tf
from tensorflow.keras import layers
import numpy as np
import matplotlib.pyplot as plt
from natsort import natsorted
class KashmirDataset:
def __init__(self,
root,
split,
num_classes=2,
crop_size=(384, 512),
mean=[0.44669862, 0.4538911, 0.42983834],
std=[0.26438654, 0.256789, 0.26210858],
label_mapping={}
):
self.root = root
self.split = split
self.num_classes = num_classes
self.img_list = natsorted(os.listdir(os.path.join(root, 'images', self.split)))
self.crop_size = crop_size
self.mean = mean
self.std = std
self.label_mapping = label_mapping
def __len__(self):
return len(self.img_list)
def read_image_and_label(self, image_path, label_path):
image = tf.io.read_file(image_path)
image = tf.image.decode_jpeg(image, channels=3)
image = tf.image.resize(image, self.crop_size, method='bilinear')
image = (image / 255.0 - self.mean) / self.std
image = tf.cast(image, tf.float32)
label = tf.io.read_file(label_path)
label = tf.image.decode_png(label, channels=1)
label = tf.image.resize(label, self.crop_size, method='nearest')
label = self.convert_label(label)
label = tf.cast(label, tf.int32)
return image, label
def convert_label(self, label):
# label = tf.where(label > 0, tf.constant(255, dtype=tf.uint8), label) # TODO: special for ITRI dataset REMOVE IT FOR OTHER DATASETS
label = tf.cast(label, tf.int32)
for k, v in self.label_mapping.items():
label = tf.where(label == k, tf.constant(v, dtype=tf.int32), label)
return label
def preprocess_data(self, image_path, label_path):
image, label = self.read_image_and_label(image_path, label_path)
return image, label
def map_function(self, image_path, label_path):
return self.preprocess_data(image_path, label_path)
def get_dataset(self):
file_paths = [(os.path.join(self.root, 'images', self.split, img_name),
os.path.join(self.root, 'labels', self.split, img_name.split('.')[0] + ".png"))
for img_name in self.img_list]
dataset = tf.data.Dataset.from_tensor_slices(file_paths)
dataset = dataset.map(lambda x: self.map_function(x[0], x[1]), num_parallel_calls=tf.data.experimental.AUTOTUNE)
return dataset
def plot_samples(images, labels, mean, std, num_classes, show_binary=True, num_samples=3):
mean = np.array(mean)
std = np.array(std)
plt.figure(figsize=(10, 5 * num_samples))
for i in range(num_samples):
img = images[i] * std + mean
img = np.clip(img * 255.0, 0, 255).astype(np.uint8)
plt.subplot(num_samples, 2, 2 * i + 1)
plt.imshow(img)
plt.title('Image')
plt.axis('off')
plt.subplot(num_samples, 2, 2 * i + 2)
if show_binary:
plt.imshow(labels[i], cmap='gray')
else:
cmap = plt.get_cmap('tab10' if num_classes <= 10 else 'tab20')
plt.imshow(labels[i], cmap=cmap)
plt.title('Label')
plt.axis('off')
plt.show()
class CityscapesDataset:
def __init__(self,
root,
split,
num_classes=19, # Cityscapes has 19 main classes
crop_size=(512, 1024),
mean=[0.28689554, 0.32513303, 0.28389177],
std=[0.18696375, 0.19017339, 0.18720214],
label_mapping={}
):
self.root = root
self.split = split
self.num_classes = num_classes
self.crop_size = crop_size
self.mean = mean
self.std = std
self.label_mapping = {
-1: 0, 0: 0, 1: 0, 2: 0, 3: 0, 4: 0, 5: 0, 6: 0,
7: 0, 8: 1, 9: 0, 10: 0, 11: 2, 12: 3, 13: 4, 14: 0,
15: 0, 16: 0, 17: 5, 18: 0, 19: 6, 20: 7, 21: 8, 22: 9,
23: 10, 24: 11, 25: 12, 26: 13, 27: 14, 28: 15, 29: 0,
30: 0, 31: 16, 32: 17, 33: 18
}
self.img_list = self.get_image_list()
self.rgb_augmentation = tf.keras.Sequential([
layers.RandomBrightness(factor=0.2),
layers.RandomContrast(factor=0.2),
])
self.rescale = tf.keras.Sequential([
layers.Rescaling(scale=1./255),
])
self.common_augmentation = tf.keras.Sequential([
layers.RandomFlip(mode='horizontal'),
layers.RandomRotation(factor=0.3),
layers.RandomTranslation(height_factor=0.2, width_factor=0.2),
])
def get_image_list(self):
img_list = []
img_dir = os.path.join(self.root, 'leftImg8bit', self.split)
for city in os.listdir(img_dir):
city_dir = os.path.join(img_dir, city)
for img_name in os.listdir(city_dir):
if img_name.endswith('_leftImg8bit.png'):
img_list.append(os.path.join(city_dir, img_name))
return natsorted(img_list)
def __len__(self):
return len(self.img_list)
def read_image_and_label(self, image_path, label_path):
image = tf.io.read_file(image_path)
image = tf.image.decode_png(image, channels=3)
image = tf.image.resize(image, self.crop_size, method='nearest')
label = tf.io.read_file(label_path)
label = tf.image.decode_png(label, channels=1)
label = tf.image.resize(label, self.crop_size, method='nearest')
label = self.convert_label(label)
label = tf.cast(label, tf.int32)
return image, label
def convert_label(self, label):
label = tf.cast(label, tf.int32)
for k, v in self.label_mapping.items():
label = tf.where(label == k, tf.constant(v, dtype=tf.int32), label)
return label
def preprocess_data(self, image_path, label_path):
image, label = self.read_image_and_label(image_path, label_path)
if self.split == 'train':
# Concatenate image and label for consistent augmentation
combined = tf.concat([tf.cast(image, tf.float32), tf.cast(label, tf.float32)], axis=-1)
combined = self.common_augmentation(combined)
image, label = combined[..., :3], combined[..., 3:]
image = self.rgb_augmentation(image)
label = tf.cast(label, tf.int32)
# Standardize image
image = self.rescale(image)
image = (image - self.mean) / self.std
return image, label
def map_function(self, image_path, label_path):
return self.preprocess_data(image_path, label_path)
def get_dataset(self):
file_paths = [(img_path, img_path.replace('leftImg8bit', 'gtFine').replace('_gtFine.png', '_gtFine_labelIds.png'))
for img_path in self.img_list]
dataset = tf.data.Dataset.from_tensor_slices(file_paths)
dataset = dataset.map(lambda x: self.map_function(x[0], x[1]), num_parallel_calls=tf.data.experimental.AUTOTUNE)
return dataset
if __name__ == "__main__":
# Usage CR_WUR Dataset
root = '/home/e300/mahmood/datasets/cr_wur_ds/'
label_map = { 0:0, 1:1, 2:2, 3:3, 4:4, 5:5, 6:6 }
mean = [0.44669862, 0.4538911, 0.42983834]
std = [0.26438654, 0.256789, 0.26210858]
split = 'train' # or 'val', 'test'
num_classes = 7 # specify the number of classes
show_binary = False # set to True to show binary images
# Usage ITRI Insulator Dataset
# root = '/home/e300/mahmood/datasets/v1_itri_formatted'
# label_map = { 0:0, 255:1, 1:1 }
# mean = [0.44669862, 0.4538911, 0.42983834]
# std = [0.26438654, 0.256789, 0.26210858]
# split = 'train' # or 'val', 'test'
# num_classes = 2 # specify the number of classes
# show_binary = False # set to True to show binary images
dataset_tf = KashmirDataset(root, split, num_classes=num_classes, mean=mean, std=std, label_mapping=label_map)
tf_dataset = dataset_tf.get_dataset()
tf_dataset = tf_dataset.batch(3)
# Take a batch from the dataset
for images, labels in tf_dataset.take(3):
print(images.shape, labels.shape)
# Convert TensorFlow tensors to NumPy arrays for plotting
images_np = images.numpy()
labels_np = labels.numpy()
print(images_np.shape, np.max(images_np), np.min(images_np))
print(labels_np.shape, np.max(labels_np), np.min(labels_np))
# Plot the samples
plot_samples(images_np, labels_np, mean, std, num_classes, show_binary)
# root = '/home/e300/mahmood/datasets/cityscapes'
# label_map = {i: i for i in range(34)} # Update label mapping if needed
# mean = [0.28805044, 0.32632137, 0.2854135 ]
# std = [0.17725339, 0.18182704, 0.17837277]
# split = 'train' # or 'val', 'test'
# num_classes = 19 # specify the number of classes
# show_binary = False # set to True to show binary images
# dataset_tf = CityscapesDataset(root, split, num_classes=num_classes, mean=mean, std=std, label_mapping=label_map)
# tf_dataset = dataset_tf.get_dataset()
# tf_dataset = tf_dataset.batch(3)
# # Take a batch from the dataset
# for images, labels in tf_dataset.take(3):
# print(images.shape, labels.shape)
# # Convert TensorFlow tensors to NumPy arrays for plotting
# images_np = images.numpy()
# labels_np = labels.numpy()
# print(np.unique(labels_np))
# print(images_np.shape, np.max(images_np), np.min(images_np))
# print(labels_np.shape, np.max(labels_np), np.min(labels_np))
# # Plot the samples
# plot_samples(images_np, labels_np, mean, std, num_classes, show_binary)