for unit in range(autoenc.encoding_layers[-1].get_shape()[1]):
slider = QtGui.QSlider(win)
slider.setOrientation(QtCore.Qt.Horizontal)
slider.setRange(-30, 200)
slider.sliderMoved.connect(sliderModev)
sliders.append(slider)
win.layout().addWidget(slider)
Add buttons to load some examples
if FLAGS.dataset.lower() == "mnist":
mnist = input_data.read_data_sets('MNIST_data', one_hot=False)
elif FLAGS.dataset.lower() == "smiley":
mnist = smiley.read_data_sets()
else:
assert False
rawdatas = []
class QButton(QtGui.QWidget):
def __init__(self, raw_image, parent=None):
QtGui.QWidget.__init__(self, parent)
self.button = QtGui.QPushButton('', self)
self.size = int(math.sqrt(raw_image.shape[1]))
self.name='me'
self.raw_image = raw_image
data = (raw_image*255).reshape(self.size, self.size).astype(np.uint8).T
self.button.clicked.connect(self.propImage)
img = QtGui.QImage(data.shape[0], data.shape[1], QtGui.QImage.Format_RGB32)
imagebox = offsetbox.AnnotationBbox( offsetbox.OffsetImage(clustered_images[i][representative].reshape(self.image_dim, self.image_dim), cmap=plt.cm.gray_r), clusters[i][representative])
ax = plt.axes(frameon=False)
ax.add_artist(imagebox)
plt.scatter(x, y, c=c, marker="x")
plt.show()
if __name__=="__main__":
data = None
num_of_classes = 0
embedded_file_path = ""
remove_arguments = list()
for i, argument in enumerate(sys.argv):
if argument == "--smiley":
data = smiley.read_data_sets()
num_of_classes = 4
remove_arguments.append(sys.argv[i])
elif argument == "--mnist":
data = input_data.read_data_sets('MNIST_data', one_hot=False)
num_of_classes = 10
remove_arguments.append(sys.argv[i])
elif argument == "--embedded_file_path":
embedded_file_path = sys.argv[i + 1]
remove_arguments.append(sys.argv[i])
remove_arguments.append(sys.argv[i + 1])
elif argument == "--model_file_path":
dir_path = os.path.dirname(sys.argv[i + 1])
model_filename = os.path.basename(os.path.normpath(sys.argv[i + 1]))
remove_arguments.append(sys.argv[i])
remove_arguments.append(sys.argv[i + 1])
# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np
from tensorflow.examples.tutorials.mnist import input_data
import smiley
import time
import autoenc_params
FLAGS = tf.app.flags.FLAGS
if __name__ == "__main__":
# Prepare the training data
if FLAGS.dataset.lower() == "mnist":
traindata = input_data.read_data_sets('MNIST_data', one_hot=True)
elif FLAGS.dataset.lower() == "smiley":
traindata = smiley.read_data_sets(one_hot=True)
else:
raise ValueError("Unknown dataset: {}".format(FLAGS.dataset))
sess, global_step, train_step, autoenc, loss, writer, saver, model_filename, cheap_summaries, expensive_summaries, test_loss_summary, classifier_out, y = autoenc_params.setupModelFromFlags()
last_cheap_summaries = 0
last_expensive_summaries = 0
test_batch = traindata.test.next_batch(FLAGS.batchsize)[0]
with sess.as_default():
try:
n_samples = traindata.train.images.shape[0]
n_epochs = FLAGS.maxepochs
batch_size = FLAGS.batchsize
n_batches = n_samples // batch_size
for epochnr in range(n_epochs):
for batchnr in range(n_batches):
