def read_data_sets(mode):
class DataSets(object):
pass
data_sets = DataSets()
train_path = './Final_Training/Images'
# test_path = './Online-Test-sort'
test_path = './Final_Test/Images'
if mode == 'train':
train_images, train_labels = readTrafficSigns.readTrafficSigns(
train_path, 'train')
data_set = DataSet(train_images, train_labels)
data_set.shuffle()
x = data_set.images
y = data_set.labels
data_sets.validation = DataSet(x[:4000], y[:4000])
data_sets.train = DataSet(x[4000:], y[4000:])
# data_sets.train.shuffle()
else:
test_images, test_labels = readTrafficSigns.readTrafficSigns(
test_path, 'test')
data_sets.test = DataSet(test_images, test_labels)
# data_sets.test.shuffle()
return data_sets
def main(self):
images, ys = readTrafficSigns.readTrafficSigns(
'D:/tf/BelgiumTSC_Training/Training')
images = np.array(images)
#强行设为整百，回头要修改！
sum = len(images) - len(images) % 100
images = images[0:sum]
ys = ys[0:sum]
ys_int = []
for a in ys:
a = int(a)
ys_int.append(a)
#下面表示拼接向量
labels_expend = tf.expand_dims(ys_int, axis=1) # 对矩阵或者向量增加一维，使其看得更清楚
index_expends = tf.expand_dims(tf.range(len(ys_int)),
axis=1) # 和labels_expend上的元素一一对应，确定横坐标
concat_result = tf.concat(values=[index_expends, labels_expend],
axis=1) # axis=1表示在列上跨越
labels_one_hot = tf.sparse_to_dense(
sparse_indices=concat_result,
output_shape=tf.zeros([len(ys), traffic_inference.OUTPUT_NODE]).shape,
sparse_values=1.0,
default_value=0.0)
train(images, labels_one_hot)
print(len(ys))
def main(argv=None):
images,ys=readTrafficSigns.readTrafficSigns('D:/files/比赛数据/交通标志识别数据/GTSRB/GTSRB/Final_Test/Images')
images = np.array(images)
ys_int = []
for a in ys:
a = int(a)
ys_int.append(a)
labels_expend = tf.expand_dims(ys_int, axis=1) # 对矩阵或者向量增加一维，使其看得更清楚
index_expends = tf.expand_dims(tf.range(len(ys_int)), axis=1) # 和labels_expend上的元素一一对应，确定横坐标
concat_result = tf.concat(values=[index_expends, labels_expend], axis=1) # axis=1表示在列上跨越
labels_one_hot = tf.sparse_to_dense(sparse_indices=concat_result,
output_shape=tf.zeros([len(ys), traffic_inference.OUTPUT_NODE]).shape,
sparse_values=1.0, default_value=0.0)
evaluate(images,labels_one_hot)
def main(self):
images, ys = readTrafficSigns.readTrafficSigns('C:/Users/曹宇/Desktop/当前识别代码/GTSRB/Final_Training/Images')
images=np.array(images)
ys_int = []
for a in ys:
a = int(a)
ys_int.append(a)
labels_expend = tf.expand_dims(ys_int, axis=1) # 对矩阵或者向量增加一维，使其看得更清楚
index_expends = tf.expand_dims(tf.range(len(ys_int)), axis=1) # 和labels_expend上的元素一一对应，确定横坐标
concat_result = tf.concat(values=[index_expends, labels_expend], axis=1) # axis=1表示在列上跨越
labels_one_hot = tf.sparse_to_dense(sparse_indices=concat_result,
output_shape=tf.zeros([len(ys), traffic_inference.OUTPUT_NODE]).shape,
sparse_values=1.0, default_value=0.0)
train(images,labels_one_hot)
print(len(ys))
def resize_and_format(image_path, width, normalise=True):
Xraw, Yraw = readTrafficSigns(image_path)
X = np.zeros([len(Yraw), Xraw[0].shape[-1], width, width],
dtype=np.float32)
Y = np.zeros(len(Yraw), dtype=np.int32)
for i, xi, yi in zip(range(len(Yraw)), Xraw, Yraw):
y = int(yi)
im = Image.fromarray(xi)
orig_size = im.size
ratio = float(width) / max(orig_size)
new_size = (int(round(ratio * orig_size[0])),
int(round(ratio * orig_size[1])))
xi = np.transpose(np.asarray(im.resize(new_size)), [2, 0, 1])
if normalise:
xi = 2 * (xi / 255.0) - 1
x = square_image(xi)
X[i] = x
Y[i] = y
return X, Y
if __name__ == "__main__":
import numpy as np
array = np.zeros((32, 32))
print(array.shape)
array = array.reshape(1, 32, 32, 1)
print(array.shape)
from readTrafficSigns import readTrafficSigns
from matplotlib import pyplot as plt
trainImages, trainLabels = readTrafficSigns(
"./traffic-signs-data/GTSRB_Final_Training_Images/GTSRB/Final_Training/Images/"
)
print(f"Labels {len(trainLabels)} and Images {len(trainImages)}")
plt.imshow(trainImages[42])
plt.show()
import numpy as np import scipy.misc import readTrafficSigns import rename_image_file rootpath = '/home/exx/' readTrafficSigns.readTrafficSigns(rootpath)
import matplotlib
import skimage.data
import skimage.transform
import matplotlib.pyplot as plt
import tensorflow as tf
from readTrafficSigns import readTrafficSigns
def display_images_and_labels(images, labels):
"""Display the first image of each label."""
unique_labels = set(labels)
plt.figure(figsize=(12,12))
i = 1
for label in unique_labels:
# Pick the first image for each label.
image = images[labels.index(label)]
plt.subplot(7, 7, i) # A grid of 8 rows x 8 columns
plt.axis('off')
plt.title("Label {0} ({1})".format(label, labels.count(label)))
i += 1
_ = plt.imshow(image)
plt.show()
images,labels=readTrafficSigns('D:/tf\BelgiumTSC_Training/Training')
for image in images[:5]:
print("shape:{0},min:{1},max:{2}".format(image.shape,image.min(),image.max()))
images32=[skimage.transform.resize(image,(12,12),mode='constant')
for image in images]
for image in images32[:5]:
print("shape:{0},min:{1},max:{2}".format(image.shape,image.min(),image.max()))
display_images_and_labels(images,labels)
import torch.utils.data as utils
from readTrafficSigns import readTrafficSigns
import random
import math
import scipy.misc
from conv_layers import ConvNet
#print (len(X_train)) #vector of dimension 39209 --> one element per image : for each image (element) we get a matrix for each pixel containing the RGBs
#print (len(y_train)) #vector of dimension 39209 --> one element per image : for each image (element) we get the number of the class it belongs to
#print (y_train[0])
#print (type(y_train))
#for the moment, images are classed in order --> firt 210 images belong to class 0 (y=0), next ones belong to class 1 etc...
# Load the raw traffic signs data
trafficSigns_train = './data/train/Final_Training/Images'
X, y = readTrafficSigns(trafficSigns_train)
X_resized = []
for i in range(len(X)):
image_resized = scipy.misc.imresize(X[i], (47, 47))
X_resized.append(image_resized)
ind = list()
# subsample the data
for i in range(0, 43):
index_value = list()
for j in range(0, 39209):
if int(y[j]) == i:
index_value.append(j)
ind.append(
random.sample(range(min(index_value),
import matplotlib
import skimage.data
import skimage.transform
import matplotlib.pyplot as plt
import tensorflow as tf
from readTrafficSigns import readTrafficSigns
def display_images_and_labels(images, labels):
"""Display the first image of each label."""
unique_labels = set(labels)
plt.figure(figsize=(12,12))
i = 1
for label in unique_labels:
# Pick the first image for each label.
image = images[labels.index(label)]
plt.subplot(7, 7, i) # A grid of 8 rows x 8 columns
plt.axis('off')
plt.title("Label {0} ({1})".format(label, labels.count(label)))
i += 1
_ = plt.imshow(image)
plt.show()
images,labels=readTrafficSigns('D:/files/比赛数据/交通标志识别数据/GTSRB/Final_Training/Images')
for image in images[:5]:
print("shape:{0},min:{1},max:{2}".format(image.shape,image.min(),image.max()))
images32=[skimage.transform.resize(image,(12,12),mode='constant')
for image in images]
for image in images32[:5]:
print("shape:{0},min:{1},max:{2}".format(image.shape,image.min(),image.max()))
display_images_and_labels(images,labels)
import tensorflow as tf
import numpy as np
import argparse
import os
from readTrafficSigns import readTrafficSigns
parser = argparse.ArgumentParser()
parser.add_argument("-i", "--index", default=100, type=int)
parser.add_argument("--path", default=os.path.join("/home", "jsmith", "GTSRB"))
parser.add_argument("--width", type=int, default=50)
parser.add_argument("--height", type=int, default=50)
args = parser.parse_args()
x_train, y_train, x_test, y_test = readTrafficSigns(args.path,
(args.width, args.height))
num_classes = len(np.unique(y_train))
x_train = x_train.astype('float32')
x_train /= 255.0
y_train = tf.keras.utils.to_categorical(y_train, num_classes)
#y_train = y_train.astype(np.int32)
x_init = x_train[args.index:args.index + 1]
y_init = y_train[args.index:args.index + 1]
file_name_x = 'gtsrb_{}.pb'.format(args.index)
file_name_y = 'gtsrb_{}_label.pb'.format(args.index)
dest='accumulate',
action='store_const',
const=sum,
default=max,
help='sum the integers (default: find the max)')
args = parser.parse_args()
def vprint(string):
if args.verbose:
print(string)
vprint('begin loading images from file')
images, labels = readTrafficSigns.readTrafficSigns(args.data)
vprint('\tdone\nplot exemplary images')
fig = plt.figure(figsize=(20, 5))
for i in range(30):
ax = plt.subplot(3, 10, i + 1)
rand_sample = random.randrange(0, len(images))
plt.imshow(images[rand_sample], cmap='gray')
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
plt.show()
vprint('preprocess input data')
labels = [int(l) for l in labels]
labels = utils.preprocess_labels(labels, size=max(labels))
def display_images_and_labels(images, labels):
"""Display the first image of each label."""
unique_labels = set(labels)
plt.figure(figsize=(12, 12))
i = 1
for label in unique_labels:
# Pick the first image for each label.
image = images[labels.index(label)]
plt.subplot(7, 7, i) # A grid of 8 rows x 8 columns
plt.axis('off')
plt.title("Label {0} ({1})".format(label, labels.count(label)))
i += 1
_ = plt.imshow(image)
plt.show()
images, labels = readTrafficSigns(
'C:/Users/曹宇/Desktop/当前识别代码/GTSRB/Final_Training/Images')
for image in images[:5]:
print("shape:{0},min:{1},max:{2}".format(image.shape, image.min(),
image.max()))
images32 = [
skimage.transform.resize(image, (12, 12), mode='constant')
for image in images
]
for image in images32[:5]:
print("shape:{0},min:{1},max:{2}".format(image.shape, image.min(),
image.max()))
display_images_and_labels(images, labels)