def main():
print('Building model...')
model = ImageModel()
print('Loading ImageNet...')
data = load_tiny_imagenet("./data")
X_train = data['X_train']
y_train = data['y_train']
X_val = data['X_val']
y_val = data['y_val']
X_test = data['X_test']
class_names = data['class_names']
test_files = data['test_files']
# y_test = data['y_test']
# print('Test labels shape: ', y_test.shape)
# permute the data axes
X_train = np.transpose(X_train, (0, 2, 3, 1))
X_val = np.transpose(X_val, (0, 2, 3, 1))
X_test = np.transpose(X_test, (0, 2, 3, 1))
print('Train data shape: ', X_train.shape)
print('Train labels shape: ', y_train.shape)
print('Validation data shape: ', X_val.shape)
print('Validation labels shape: ', y_val.shape)
print('Test data shape: ', X_test.shape)
print(len(class_names))
sess = tf.Session()
sess.run(tf.global_variables_initializer())
cp_saver = tf.train.Saver(tf.trainable_variables())
# cp_saver = tf.train.Saver()
cp_saver.restore(sess, "ckpts/v_simple/model.ckpt-10")
# print('Training')
# # logs_path = "tensorboard/" + strftime("%Y_%m_%d_%H_%M_%S", gmtime())
# # train_writer = tf.summary.FileWriter(logs_path + '/train', session.graph)
# model.run(sess, X_train, y_train, True, epochs=10)
# print('Validation')
# model.run(sess, X_val, y_val, False, epochs=1)
# # run_model(sess,X_val,y_val)
model.run_with_valid(sess,
X_train,
y_train,
X_val,
y_val,
epochs=30,
batch_size=32)
def load_dataset(dataset):
if dataset == "cifar10":
return generate_dataset_cifar10("../datasets/cifar-10-batches-py")
elif dataset == "tiny200":
class_names, X_train, y_train, X_val, y_val = load_tiny_imagenet(
path="../datasets/tiny-imagenet-200/",
wnids_path="",
resize='False',
num_classes=200,
dtype=np.float32)
N = X_train.shape[0]
Y_train = np.zeros((N, 200))
Y_train[np.arange(N), y_train] = 1
N = X_val.shape[0]
Y_val = np.zeros((N, 200))
Y_val[np.arange(N), y_val] = 1
return X_train, Y_train, X_val, Y_val
def process_images(wnids_path='', resize='False', num_classes=200, normalize='False'):
# Path to tiny imagenet dataset
if wnids_path == '':
wnids_path = input('Enter the relative path to the directory containing the wnids/words files from sets/: ')
wnids_path = os.path.join('..', 'sets', wnids_path)
# Generate data fields - test data has no labels so ignore it
classes, x_train, y_train, x_val, y_val = load_tiny_imagenet(os.path.join('tiny-imagenet-200'), wnids_path, num_classes=num_classes, resize=resize)
# Format data to be the correct shape
x_train = np.einsum('iljk->ijkl', x_train)
x_val = np.einsum('iljk->ijkl', x_val)
if normalize.lower() == 'true':
x_train /= 255
x_val /= 255
# Convert labels to one hot vectors
y_train = keras.utils.to_categorical(y_train, num_classes)
y_val = keras.utils.to_categorical(y_val, num_classes)
return x_train, y_train, x_val, y_val, wnids_path
def process_images(num_classes=10):
# Path to tiny imagenet dataset
#path = input('Enter the relative path to the directory containing the wnids/words files: ')
path = os.path.join('tiny-imagenet-200')
#path = os.path.join('tiny-imagenet-200', 'random', '0')
print(path)
# Generate data fields - test data has no labels so ignore it
classes, x_train, y_train, x_val, y_val = load_tiny_imagenet(
path,
os.path.join('random', '0'),
num_classes=num_classes,
resize=True)
# Get number of classes specified in order from [0, num_classes)
print(classes)
#print(x_train)
print(x_train.shape)
print(y_train.shape)
"""
if num_classes > 200:
print('Set number of classes to maximum of 200\n')
num_classes = 200
elif num_classes != 200:
train_indices = [index for index, label in enumerate(y_train) if label < num_classes]
val_indices = [index for index, label in enumerate(y_val) if label < num_classes]
x_train = x_train[train_indices]
y_train = y_train[train_indices]
x_val = x_val[val_indices]
y_val = y_val[val_indices]
"""
# Format data to be the correct shape
x_train = np.einsum('iljk->ijkl', x_train)
x_val = np.einsum('iljk->ijkl', x_val)
# Convert labels to one hot vectors
y_train = keras.utils.to_categorical(y_train, num_classes)
y_val = keras.utils.to_categorical(y_val, num_classes)
return x_train, y_train, x_val, y_val
def main(argv):
#%% load data
data = load_tiny_imagenet('datasets/tiny-imagenet-100-A',
subtract_mean=True)
print data['X_train'].shape
mnist_graph = tf.Graph()
with mnist_graph.as_default():
# Generate placeholders for the images and labels.
images_placeholder = tf.placeholder(tf.float32)
labels_placeholder = tf.placeholder(tf.int32)
tf.add_to_collection("images", images_placeholder) # Remember this Op.
tf.add_to_collection("labels", labels_placeholder) # Remember this Op.
# Build a Graph that computes predictions from the inference model.
logits, keep_prob = mnist_inference(images_placeholder,
FLAGS.HIDDEN1_UNITS,
FLAGS.HIDDEN2_UNITS)
tf.add_to_collection("logits", logits) # Remember this Op.
# Add to the Graph the Ops that calculate and apply gradients.
train_op, loss_op = mnist_training(logits, labels_placeholder,
FLAGS.lr)
# Add the variable initializer Op.
init = tf.global_variables_initializer()
# Create a saver for writing training checkpoints.
saver = tf.train.Saver()
tf.summary.scalar("Cost", loss_op)
# Uncomment the following line to see what we have constructed.
# tf.train.write_graph(tf.get_default_graph().as_graph_def(),
# "/tmp", "complete.pbtxt", as_text=True)
with tf.Session(graph=mnist_graph) as sess:
# Merge all the tf.summary
summary_op = tf.summary.merge_all()
train_writer = tf.summary.FileWriter('model', sess.graph)
# Run the Op to initialize the variables.
sess.run(init)
# Start the training loop.
train_ds = DataSet(data['X_train'], data['y_train'])
max_steps = int(FLAGS.MAX_STEPS)
for step in xrange(max_steps):
images, labels = train_ds.next_batch(256)
# Run one step of the model. The return values are the activations
# from the `train_op` (which is discarded) and the `loss` Op. To
# inspect the values of your Ops or variables, you may include them
# in the list passed to sess.run() and the value tensors will be
# returned in the tuple from the call.
_, loss, summary = sess.run(
[train_op, loss_op, summary_op],
feed_dict={
images_placeholder: images,
labels_placeholder: labels,
keep_prob: 1.0
})
train_writer.add_summary(summary, step)
if step % 100 == 0: # Record execution stats
print('Step %d/%d -- loss: %f' % (step, max_steps, loss))
# Write a checkpoint.
train_writer.close()
checkpoint_file = os.path.join(FLAGS.MODEL_SAVE_PATH, 'checkpoint')
saver.save(sess, checkpoint_file, global_step=step)
print train_ds.epoch_completed
CURRENT_DIR = os.path.dirname(os.path.abspath(__file__))
sys.path.insert(0, CURRENT_DIR + '\\helper')
from data_utils import load_tiny_imagenet
##from image_utils import image_from_url
##
sys.path.insert(0, CURRENT_DIR + '\\classifier')
from pretrained_cnn import PretrainedCNN
plt.rcParams['figure.figsize'] = (10.0, 8.0) # set default size of plots
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'
data = load_tiny_imagenet('datasets/tiny-imagenet-100-A', subtract_mean=True)
# Visualize some examples of the training data
classes_to_show = 7
examples_per_class = 5
class_idxs = np.random.choice(len(data['class_names']),
size=classes_to_show,
replace=False)
for i, class_idx in enumerate(class_idxs):
train_idxs, = np.nonzero(data['y_train'] == class_idx)
train_idxs = np.random.choice(train_idxs,
size=examples_per_class,
replace=False)
for j, train_idx in enumerate(train_idxs):
img = deprocess_image(data['X_train'][train_idx], data['mean_image'])
