-
Notifications
You must be signed in to change notification settings - Fork 33
/
helper_cityscapes.py
192 lines (169 loc) · 7.45 KB
/
helper_cityscapes.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
import re
import cv2
import random
import numpy as np
import os.path
import scipy.misc
import shutil
import zipfile
import time
import tensorflow as tf
from scipy import ndimage
from glob import glob
from urllib.request import urlretrieve
from tqdm import tqdm
class DLProgress(tqdm):
last_block = 0
def hook(self, block_num=1, block_size=1, total_size=None):
self.total = total_size
self.update((block_num - self.last_block) * block_size)
self.last_block = block_num
def maybe_download_pretrained_vgg(data_dir):
"""
Download and extract pretrained vgg model if it doesn't exist
:param data_dir: Directory to download the model to
"""
vgg_filename = 'vgg.zip'
vgg_path = os.path.join(data_dir, 'vgg')
vgg_files = [
os.path.join(vgg_path, 'variables/variables.data-00000-of-00001'),
os.path.join(vgg_path, 'variables/variables.index'),
os.path.join(vgg_path, 'saved_model.pb')]
missing_vgg_files = [vgg_file for vgg_file in vgg_files if not os.path.exists(vgg_file)]
if missing_vgg_files:
# Clean vgg dir
if os.path.exists(vgg_path):
shutil.rmtree(vgg_path)
os.makedirs(vgg_path)
# Download vgg
print('Downloading pre-trained vgg model...')
with DLProgress(unit='B', unit_scale=True, miniters=1) as pbar:
urlretrieve(
'https://s3-us-west-1.amazonaws.com/udacity-selfdrivingcar/vgg.zip',
os.path.join(vgg_path, vgg_filename),
pbar.hook)
# Extract vgg
print('Extracting model...')
zip_ref = zipfile.ZipFile(os.path.join(vgg_path, vgg_filename), 'r')
zip_ref.extractall(data_dir)
zip_ref.close()
# Remove zip file to save space
os.remove(os.path.join(vgg_path, vgg_filename))
def img_size(img):
return (img.shape[0], img.shape[1])
def random_crop(img, gt):
h,w = img_size(img)
nw = random.randint(768, w-2) # Random crop size
nh = int(nw / 2) # Keep original aspect ration
x1 = random.randint(0, w - nw) # Random position of crop
y1 = random.randint(0, h - nh)
return img[y1:(y1+nh), x1:(x1+nw), :], gt[y1:(y1+nh), x1:(x1+nw), :]
def bc_img(img, s = 1.0, m = 0.0):
img = img.astype(np.int)
img = img * s + m
img[img > 255] = 255
img[img < 0] = 0
img = img.astype(np.uint8)
return img
train_dataset_dir = 'data/leftImg8bit/train_ds/'
val_dataset_dir = 'data/leftImg8bit/val_ds/'
gt_dataset_dir = 'data/leftImg8bit/gt_ds/'
road_color = np.array([128, 64, 128, 255])
car_color = np.array([0, 0, 142, 255])
img_input = []
gt_input = []
#Load data into RAM, if you have enough space
def load_data():
image_paths = os.listdir(gt_dataset_dir)
for image_file in image_paths:
gt_image = cv2.imread(os.path.join(gt_dataset_dir, image_file))
image = cv2.imread(os.path.join(train_dataset_dir, image_file[:-5]+'.png'))
gt_input.append(gt_image)
img_input.append(image)
def gen_batch_function(data_folder, image_shape):
"""
Generate function to create batches of training data
:param data_folder: Path to folder that contains all the datasets
:param image_shape: Tuple - Shape of image
:return:
"""
def get_batches_fn(batch_size):
"""
Create batches of training data
:param batch_size: Batch Size
:return: Batches of training data
"""
image_paths = os.listdir(gt_dataset_dir)
random.shuffle(image_paths)
for batch_i in range(0, len(image_paths), batch_size):
images = []
gt_images = []
for image_file in image_paths[batch_i:batch_i+batch_size]:
gt_image = cv2.imread(os.path.join(gt_dataset_dir, image_file))
image = cv2.imread(os.path.join(train_dataset_dir, image_file[:-5]+'.png'))
image, gt_image = random_crop(image, gt_image) #Random crop augmentation
image = cv2.resize(image, image_shape)
gt_image = cv2.resize(gt_image, image_shape)
contr = random.uniform(0.85, 1.15) # Contrast augmentation
bright = random.randint(-40, 30) # Brightness augmentation
image = bc_img(image, contr, bright)
gt_road = np.all(gt_image == road_color, axis=2)
gt_road = gt_road.reshape(*gt_road.shape, 1)
gt_car = np.all(gt_image == car_color, axis=2)
gt_car = gt_car.reshape(*gt_car.shape, 1)
gt_obj = np.concatenate((gt_road, gt_car), axis=2)
gt_bg = np.all(gt_obj == 0, axis=2)
gt_bg = gt_bg.reshape(*gt_bg.shape, 1)
images.append(image)
gt_images.append(gt_image)
yield np.array(images), np.array(gt_images)
return get_batches_fn
def denoise_img(img):
eroded_img = ndimage.binary_erosion(img)
return ndimage.binary_propagation(eroded_img, mask=img)
def gen_test_output(sess, logits, keep_prob, image_pl, data_folder, image_shape, num_classes):
"""
Generate test output using the test images
:param sess: TF session
:param logits: TF Tensor for the logits
:param keep_prob: TF Placeholder for the dropout keep robability
:param image_pl: TF Placeholder for the image placeholder
:param data_folder: Path to the folder that contains the datasets
:param image_shape: Tuple - Shape of image
:return: Output for for each test image
"""
for image_file in os.listdir(val_dataset_dir):
image = scipy.misc.imresize(scipy.misc.imread(os.path.join(val_dataset_dir, image_file)), image_shape)
street_im = scipy.misc.toimage(image)
im_softmax = sess.run(
[tf.nn.softmax(logits)],
{keep_prob: 1.0, image_pl: [image]})
# Road
im_softmax_r = im_softmax[0][:, 0].reshape(image_shape[0], image_shape[1])
segmentation_r = (im_softmax_r > 0.5).reshape(image_shape[0], image_shape[1], 1)
mask = np.dot(segmentation_r, np.array([[128, 64, 128, 64]]))
mask = scipy.misc.toimage(mask, mode="RGBA")
street_im.paste(mask, box=None, mask=mask)
# Car
im_softmax_r = im_softmax[0][:, 1].reshape(image_shape[0], image_shape[1])
segmentation_r = (im_softmax_r > 0.5).reshape(image_shape[0], image_shape[1], 1)
mask = np.dot(segmentation_r, np.array([[0, 0, 142, 64]]))
mask = scipy.misc.toimage(mask, mode="RGBA")
street_im.paste(mask, box=None, mask=mask)
# Signs
#im_softmax_r = im_softmax[0][:, 3].reshape(image_shape[0], image_shape[1])
#segmentation_r = (im_softmax_r > 0.5).reshape(image_shape[0], image_shape[1], 1)
#mask += np.dot(segmentation_r, np.array([[220, 220, 0, 127]]))
yield os.path.basename(image_file), np.array(street_im)
def save_inference_samples(runs_dir, data_dir, sess, image_shape, logits, keep_prob, input_image, num_classes):
# Make folder for current run
output_dir = os.path.join(runs_dir, str(time.time()))
if os.path.exists(output_dir):
shutil.rmtree(output_dir)
os.makedirs(output_dir)
# Run NN on test images and save them to HD
print('Training Finished. Saving test images to: {}'.format(output_dir))
image_outputs = gen_test_output(
sess, logits, keep_prob, input_image, os.path.join(data_dir, 'data_road/testing'), image_shape, num_classes)
for name, image in image_outputs:
scipy.misc.imsave(os.path.join(output_dir, name), image)