Created
October 23, 2017 08:49
-
-
Save anonymous/d5f13337cb67c48f522cae96d69704f7 to your computer and use it in GitHub Desktop.
training on faces with tf
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
""" Build an Image Dataset in TensorFlow. | |
For this example, you need to make your own set of images (JPEG). | |
We will show 2 different ways to build that dataset: | |
- From a root folder, that will have a sub-folder containing images for each class | |
``` | |
ROOT_FOLDER | |
|-------- SUBFOLDER (CLASS 0) | |
| | | |
| | ----- image1.jpg | |
| | ----- image2.jpg | |
| | ----- etc... | |
| | |
|-------- SUBFOLDER (CLASS 1) | |
| | | |
| | ----- image1.jpg | |
| | ----- image2.jpg | |
| | ----- etc... | |
``` | |
- From a plain text file, that will list all images with their class ID: | |
``` | |
/path/to/image/1.jpg CLASS_ID | |
/path/to/image/2.jpg CLASS_ID | |
/path/to/image/3.jpg CLASS_ID | |
/path/to/image/4.jpg CLASS_ID | |
etc... | |
``` | |
Below, there are some parameters that you need to change (Marked 'CHANGE HERE'), | |
such as the dataset path. | |
Author: Aymeric Damien | |
Project: https://github.com/aymericdamien/TensorFlow-Examples/ | |
""" | |
from __future__ import print_function | |
import tensorflow as tf | |
import os | |
# Dataset Parameters - CHANGE HERE | |
MODE = 'folder' # or 'file', if you choose a plain text file (see above). | |
DATASET_PATH = '/media/sf_data/faces/tv10/' # the dataset file or root folder path. | |
# Image Parameters | |
N_CLASSES = 30 # CHANGE HERE, total number of classes | |
IMG_HEIGHT = 64 # CHANGE HERE, the image height to be resized to | |
IMG_WIDTH = 64 # CHANGE HERE, the image width to be resized to | |
CHANNELS = 1 # The 3 color channels, change to 1 if grayscale | |
# Reading the dataset | |
# 2 modes: 'file' or 'folder' | |
def read_images(dataset_path, mode, batch_size): | |
imagepaths, labels = list(), list() | |
if mode == 'file': | |
# Read dataset file | |
data = open(dataset_path, 'r').read().splitlines() | |
for d in data: | |
imagepaths.append(d.split(' ')[0]) | |
labels.append(int(d.split(' ')[1])) | |
elif mode == 'folder': | |
# An ID will be affected to each sub-folders by alphabetical order | |
label = 0 | |
# List the directory | |
try: # Python 2 | |
classes = sorted(os.walk(dataset_path).next()[1]) | |
except Exception: # Python 3 | |
classes = sorted(os.walk(dataset_path).__next__()[1]) | |
# List each sub-directory (the classes) | |
for c in classes: | |
c_dir = os.path.join(dataset_path, c) | |
try: # Python 2 | |
walk = os.walk(c_dir).next() | |
except Exception: # Python 3 | |
walk = os.walk(c_dir).__next__() | |
# Add each image to the training set | |
for sample in walk[2]: | |
# Only keeps jpeg images | |
if sample.endswith('.png'): | |
imagepaths.append(os.path.join(c_dir, sample)) | |
labels.append(label) | |
label += 1 | |
else: | |
raise Exception("Unknown mode.") | |
# Convert to Tensor | |
imagepaths = tf.convert_to_tensor(imagepaths, dtype=tf.string) | |
labels = tf.convert_to_tensor(labels, dtype=tf.int32) | |
# Build a TF Queue, shuffle data | |
image, label = tf.train.slice_input_producer([imagepaths, labels], | |
shuffle=True) | |
# Read images from disk | |
image = tf.read_file(image) | |
image = tf.image.decode_png(image, channels=CHANNELS) | |
# Resize images to a common size | |
image = tf.image.resize_images(image, [IMG_HEIGHT, IMG_WIDTH]) | |
# Normalize | |
image = image * 1.0/127.5 - 1.0 | |
# Create batches | |
X, Y = tf.train.batch([image, label], batch_size=batch_size, | |
capacity=batch_size * 8, | |
num_threads=4) | |
return X, Y | |
# ----------------------------------------------- | |
# THIS IS A CLASSIC CNN (see examples, section 3) | |
# ----------------------------------------------- | |
# Note that a few elements have changed (usage of queues). | |
# Parameters | |
learning_rate = 0.001 | |
num_steps = 10 | |
batch_size = 128 | |
display_step = 1 | |
# Network Parameters | |
dropout = 0.75 # Dropout, probability to keep units | |
# Build the data input | |
X, Y = read_images(DATASET_PATH, MODE, batch_size) | |
# Create model | |
def conv_net(x, n_classes, dropout, reuse, is_training): | |
# Define a scope for reusing the variables | |
with tf.variable_scope('ConvNet', reuse=reuse): | |
# Convolution Layer with 32 filters and a kernel size of 5 | |
conv1 = tf.layers.conv2d(x, 32, 5, activation=tf.nn.relu) | |
# Max Pooling (down-sampling) with strides of 2 and kernel size of 2 | |
conv1 = tf.layers.max_pooling2d(conv1, 2, 2) | |
# Convolution Layer with 32 filters and a kernel size of 5 | |
conv2 = tf.layers.conv2d(conv1, 64, 3, activation=tf.nn.relu) | |
# Max Pooling (down-sampling) with strides of 2 and kernel size of 2 | |
conv2 = tf.layers.max_pooling2d(conv2, 2, 2) | |
# Flatten the data to a 1-D vector for the fully connected layer | |
fc1 = tf.contrib.layers.flatten(conv2) | |
# Fully connected layer (in contrib folder for now) | |
fc1 = tf.layers.dense(fc1, 1024) | |
# Apply Dropout (if is_training is False, dropout is not applied) | |
fc1 = tf.layers.dropout(fc1, rate=dropout, training=is_training) | |
# Output layer, class prediction | |
out = tf.layers.dense(fc1, n_classes) | |
# Because 'softmax_cross_entropy_with_logits' already apply softmax, | |
# we only apply softmax to testing network | |
out = tf.nn.softmax(out) if not is_training else out | |
return out | |
# Because Dropout have different behavior at training and prediction time, we | |
# need to create 2 distinct computation graphs that share the same weights. | |
# Create a graph for training | |
logits_train = conv_net(X, N_CLASSES, dropout, reuse=False, is_training=True) | |
# Create another graph for testing that reuse the same weights | |
logits_test = conv_net(X, N_CLASSES, dropout, reuse=True, is_training=False) | |
# Define loss and optimizer (with train logits, for dropout to take effect) | |
loss_op = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits( | |
logits=logits_train, labels=Y)) | |
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate) | |
train_op = optimizer.minimize(loss_op) | |
# Evaluate model (with test logits, for dropout to be disabled) | |
correct_pred = tf.equal(tf.argmax(logits_test, 1), tf.cast(Y, tf.int64)) | |
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32)) | |
# Initialize the variables (i.e. assign their default value) | |
init = tf.global_variables_initializer() | |
# Saver object | |
saver = tf.train.Saver() | |
# Start training | |
with tf.Session() as sess: | |
# Run the initializer | |
sess.run(init) | |
# Start the data queue | |
tf.train.start_queue_runners() | |
# Training cycle | |
for step in range(1, num_steps+1): | |
if step % display_step == 0: | |
# Run optimization and calculate batch loss and accuracy | |
_, loss, acc = sess.run([train_op, loss_op, accuracy]) | |
print("Step " + str(step) + ", Minibatch Loss= " + \ | |
"{:.4f}".format(loss) + ", Training Accuracy= " + \ | |
"{:.3f}".format(acc)) | |
else: | |
# Only run the optimization op (backprop) | |
sess.run(train_op) | |
print("Optimization Finished!") | |
# Save your model | |
saver.save(sess, 'my_tf_model') |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment