def inputs(train=True):
"""Construct input for evaluation using the Reader ops.
Args:
eval_data: bool, indicating if one should use the train or eval data set.
Raises:
ValueError: if no data_dir
Returns:
images: Images. 3D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE] size.
labels: Labels. 1D tensor of [batch_size] size.
"""
if train:
filenames = [os.path.join(FLAGS.data_dir, 'train.csv') for _ in xrange(1000)]
else:
filenames = [os.path.join(FLAGS.data_dir, 'test1.csv'),
os.path.join(FLAGS.data_dir, 'test2.csv')]
filename_queue = tf.train.string_input_producer(filenames)
read_input = input.read(filename_queue)
reshaped_image = tf.cast(read_input.image, tf.float32)
height = IMAGE_SIZE
width = IMAGE_SIZE
# Image processing for evaluation.
# Crop the central [height, width] of the image.
resized_image = tf.image.resize_image_with_crop_or_pad(reshaped_image,
width, height)
# Subtract off the mean and divide by the variance of the pixels.
float_image = tf.image.per_image_whitening(resized_image)
# Ensure that the random shuffling has good mixing properties.
min_fraction_of_examples_in_queue = 0.4
num_examples_per_epoch = NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN if train else NUM_EXAMPLES_PER_EPOCH_FOR_EVAL
min_queue_examples = int(num_examples_per_epoch *
min_fraction_of_examples_in_queue)
# Generate a batch of images and labels by building up a queue of examples.
return _generate_image_and_label_batch(float_image, read_input.label,
min_queue_examples)
def recognizeFile(models, file, translate='', rotate='', scale=''):
"""!
Match a single file and return the resulting scores as well as the
normalization parameters.
@param models list: The previously trained HMM models
@param file String: The file containing the motion.
@param translate String: The normalization type for correcting translation
@param rotate String: The normalization type for correcting rotation
@param scale String: The normalization type for correcting scaling
@return An array of the model scores, translation, rotation, scaling parameters
"""
#print(file)
#read motion and normalize
motion = input.read(file)
motion,t,r,s = normalization.normalize(motion, translate, rotate, scale)
plot.addPlot(motion[:,1:4], file)
#writePointsToGrapherFile(motion, file)
scores = []
# check motion score (likelyhood) for each recording
for i,model in enumerate(models):
scores.append(float(model.score(motion)))
return numpy.array(scores), t, r, s
def distorted_inputs():
"""Construct distorted input for CIFAR training using the Reader ops.
Raises:
ValueError: if no data_dir
Returns:
images: Images. 3D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE] size.
labels: Labels. 1D tensor of [batch_size] size.
"""
filenames = [
os.path.join(FLAGS.data_dir, 'train.csv') for _ in xrange(10000)
]
filename_queue = tf.train.string_input_producer(filenames)
read_input = input.read(filename_queue)
reshaped_image = tf.cast(read_input.image, tf.float32)
height = IMAGE_SIZE
width = IMAGE_SIZE
# Image processing for training the network. Note the many random
# distortions applied to the image.
# Randomly crop a [height, width] section of the image.
distorted_image = tf.image.random_crop(reshaped_image, [height, width])
#distorted_image = tf.image.resize_image_with_crop_or_pad(reshaped_image,
# width, height)
# Randomly flip the image horizontally.
distorted_image = tf.image.random_flip_left_right(distorted_image)
# Because these operations are not commutative, consider randomizing
# randomize the order their operation.
#distorted_image = tf.image.random_brightness(distorted_image,
# max_delta=10)
brightness_adj = random.randint(-20, 20)
distorted_image = tf.image.adjust_brightness(distorted_image,
brightness_adj,
min_value=1,
max_value=255)
contrast_adj = random.uniform(0.85, 1.15)
distorted_image = tf.image.adjust_contrast(distorted_image,
contrast_adj,
min_value=1,
max_value=255)
#distorted_image = tf.image.random_contrast(distorted_image,
# lower=0.7, upper=1.3)
# Subtract off the mean and divide by the variance of the pixels.
float_image = tf.image.per_image_whitening(distorted_image)
# Ensure that the random shuffling has good mixing properties.
min_fraction_of_examples_in_queue = 0.4
min_queue_examples = int(NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN *
min_fraction_of_examples_in_queue)
print('Filling queue with %d images before starting to train. '
'This will take a few minutes.' % min_queue_examples)
# Generate a batch of images and labels by building up a queue of examples.
return _generate_image_and_label_batch(float_image, read_input.label,
min_queue_examples)
def distorted_inputs():
"""Construct distorted input for CIFAR training using the Reader ops.
Raises:
ValueError: if no data_dir
Returns:
images: Images. 3D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE] size.
labels: Labels. 1D tensor of [batch_size] size.
"""
filenames = [os.path.join(FLAGS.data_dir, 'train.csv') for _ in xrange(10000)]
filename_queue = tf.train.string_input_producer(filenames)
read_input = input.read(filename_queue)
reshaped_image = tf.cast(read_input.image, tf.float32)
height = IMAGE_SIZE
width = IMAGE_SIZE
# Image processing for training the network. Note the many random
# distortions applied to the image.
# Randomly crop a [height, width] section of the image.
distorted_image = tf.image.random_crop(reshaped_image, [height, width])
#distorted_image = tf.image.resize_image_with_crop_or_pad(reshaped_image,
# width, height)
# Randomly flip the image horizontally.
distorted_image = tf.image.random_flip_left_right(distorted_image)
# Because these operations are not commutative, consider randomizing
# randomize the order their operation.
#distorted_image = tf.image.random_brightness(distorted_image,
# max_delta=10)
brightness_adj = random.randint(-20, 20)
distorted_image = tf.image.adjust_brightness(distorted_image,
brightness_adj,
min_value=1,
max_value=255)
contrast_adj = random.uniform(0.85, 1.15)
distorted_image = tf.image.adjust_contrast(distorted_image,
contrast_adj,
min_value=1,
max_value=255)
#distorted_image = tf.image.random_contrast(distorted_image,
# lower=0.7, upper=1.3)
# Subtract off the mean and divide by the variance of the pixels.
float_image = tf.image.per_image_whitening(distorted_image)
# Ensure that the random shuffling has good mixing properties.
min_fraction_of_examples_in_queue = 0.4
min_queue_examples = int(NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN *
min_fraction_of_examples_in_queue)
print ('Filling queue with %d images before starting to train. '
'This will take a few minutes.' % min_queue_examples)
# Generate a batch of images and labels by building up a queue of examples.
return _generate_image_and_label_batch(float_image, read_input.label,
min_queue_examples)
def plotFile(file, translate='', rotate='', scale=''):
"""!
Read a single motion from a file and add it to the current plot list.
@param file String: The file containing the motion.
@param translate String: The normalization type for correcting translation
@param rotate String: The normalization type for correcting rotation
@param scale String: The normalization type for correcting scaling
"""
#read motion and normalize
motion = input.read(file)
motion,t,r,s = normalization.normalize(motion, translate, rotate, scale)
plot.addPlot(motion[:,1:4], file)
def readNormalized(file, translate='', rotate='', scale='', clean=True):
"""!
Read a motion from a file path and apply normalization to it.
@param file String: The file path of the motion
@param translate: The normalization for translating the motions
@param rotate: The normalization for rotating the motions
@param scale: The normalization for scaling the motions
@param cleaning: Remove duplicate points and large jumps, default true
@return: The normalized motion and the normalization parameters
"""
motion = read(file)
motion, t, r, s = normalize(motion, translate, rotate, scale, clean)
return motion, t, r, s
def createMotionsAndLengths(path, translate='', rotate='', scale=''):
"""!
Read the motions from a folder and create a concatenated array with the lengths.
The given directory 'path' must contain a subdirectory 'training' containing the motions
as individual csv files.
@param path String: The path to the motion data.
@param translate: The normalization for translating the motions
@param rotate: The normalization for rotating the motions
@param scale: The normalization for scaling the motions
@return: The concatenated motions and a list of the motion lengths
"""
# list of motions read from the file
motions = []
# length of motions read from the file
lengths = []
count = 0
plot.clearPlot()
input.logLn('\n- ' + '{:<10}'.format(path + ':'))
# read all files from directory associated with one motion
for file in sorted(glob.glob(path + '/training/*.csv')):
print(file)
input.logLn(string.basename(string.splitext(file)[0]))
count += 1
#read motion and normalize
motion = input.read(file)
motion,t,r,s = normalization.normalize(motion, translate, rotate, scale)
# Add to plot all training plots
plot.addPlot(motion[:,1:4], file)
# Add motion to the list of motions
motions.append(motion)
# Add length (number of poses in motion) to the list of lengths
lengths.append(len(motion))
_, folderName = string.split(string.dirname(path))
# Plot all training motions
plot.plot('../plots/' + folderName + ' training')
# The observations are a list of poses
X = numpy.concatenate(motions)
return X, lengths
def inputs(train=True):
"""Construct input for evaluation using the Reader ops.
Args:
eval_data: bool, indicating if one should use the train or eval data set.
Raises:
ValueError: if no data_dir
Returns:
images: Images. 3D tensor of [batch_size, IMAGE_SIZE, IMAGE_SIZE] size.
labels: Labels. 1D tensor of [batch_size] size.
"""
if train:
filenames = [
os.path.join(FLAGS.data_dir, 'train.csv') for _ in xrange(1000)
]
else:
filenames = [
os.path.join(FLAGS.data_dir, 'test1.csv'),
os.path.join(FLAGS.data_dir, 'test2.csv')
]
filename_queue = tf.train.string_input_producer(filenames)
read_input = input.read(filename_queue)
reshaped_image = tf.cast(read_input.image, tf.float32)
height = IMAGE_SIZE
width = IMAGE_SIZE
# Image processing for evaluation.
# Crop the central [height, width] of the image.
resized_image = tf.image.resize_image_with_crop_or_pad(
reshaped_image, width, height)
# Subtract off the mean and divide by the variance of the pixels.
float_image = tf.image.per_image_whitening(resized_image)
# Ensure that the random shuffling has good mixing properties.
min_fraction_of_examples_in_queue = 0.4
num_examples_per_epoch = NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN if train else NUM_EXAMPLES_PER_EPOCH_FOR_EVAL
min_queue_examples = int(num_examples_per_epoch *
min_fraction_of_examples_in_queue)
# Generate a batch of images and labels by building up a queue of examples.
return _generate_image_and_label_batch(float_image, read_input.label,
min_queue_examples)
error = sum_error(sigma)
mses.append(error)
if mses[-1] < 10:
return mses, wo, wh
sigma_o = np.dot(sigma.T, net_h) # == > 1 * 10
sigma_h = np.dot((np.dot(sigma, wo) * sigmoid_dash(net_h_active)).T, x) # == > 10 * 8
wo = wo + 0.0001 * sigma_o # == > 1 * 10
wh = wh + 0.0001 * sigma_h # == > 10 * 8
if __name__ == '__main__':
np.random.seed(1)
m, l, n, x, y = read()
# m ==> 8
# l ==> 10
# n ==> 1
wh = np.random.uniform(low=-5, high=5, size=(l, m)) # 10 * 8
wo = np.random.uniform(low=-5, high=5, size=(n, l)) # 1 * 10
x = normalize_all(x)
y = normalize(y)
mses, wo, wh = backpropagation(x, y, wo, wh)
save(wo, wh)
plt.plot(mses)
print(mses[-1])
plt.show()
flags=['multi_index'],
op_flags=['readwrite']) as it:
for cell in it:
x = it.multi_index[0]
y = it.multi_index[1]
values[x, y] = space[sorted(list(range(y, y + size)) * size),
list(range(x, x + size)) * size].sum()
return values
def part2(input):
serial_num = int(input)
space = np.empty((300, 300), dtype=np.int8)
assign_power(space, serial_num)
max_powers = np.arange(1, 300)
with np.nditer(max_powers, flags=['f_index'],
op_flags=['readwrite']) as it:
for p in it:
max_powers[it.index] = max_square(space,
max_powers[it.index]).max()
return
if __name__ == '__main__':
import input as inp
DAY = 11
input = inp.read(DAY)
print(part1(input))
# print(part2(input))
# print(part1('42'))
# import matplotlib.pyplot as plt
# from mpl_toolkits.mplot3d import Axes3D
#
import numpy
import input
import plot
import glob
import clean
#for file in sorted(glob.glob('../recordings/writing/training/*.csv')):
motion = input.read('../recordings/wiping/translated/17-09-01 16-47-01.csv')
motion, _ = clean.removeDoublePoints(motion)
newMotion, count = clean.removeInvalidParts(motion)
print('Removed ' + str(count) + ' points')
print('New length is ', len(newMotion[:, 0]))
plot.addPlot(motion[:, 1:4], "Original")
plot.addPlot(newMotion[:, 1:4], "Cleaned")
plot.plot()
#
# motion = input.read('all/17-09-01 16-44-48.csv')
# motion2 = input.read('all/17-09-01 16-44-51.csv')
#
#
# plot.addPlot(motion[:,1:4], 'Something')
# plot.addPlot(motion2[:,1:4], 'Something else')
# plot.plot()
# from os.path import basename
