def __init__(self,
w_decay,
keep_p,
optimizer,
input_shape=[None, 28, 28],
input_reshape=[-1, 28, 28, 1]):
self.img_size = inception.inception_v1.default_image_size
self.features = tf.placeholder(tf.float32, input_shape)
self.images = tf.reshape(self.features, input_reshape)
self.images = tf.image.resize_images(self.images,
[self.img_size, self.img_size])
self.labels = tf.placeholder(tf.float32, [None, 10])
self.labels = tf.cast(self.labels, tf.int32)
with tf.variable_scope("Inception") as scope:
self.train_digits, _ = inception.inception_v1(self.images,
num_classes=10,
is_training=True)
scope.reuse_variables()
self.pred_digits, _ = inception.inception_v1(self.images,
num_classes=10,
is_training=False)
self.logits = self.pred_digits
self.correct_prediction = tf.equal(tf.argmax(self.pred_digits, 1),
tf.argmax(self.labels, 1))
self.accuracy = tf.reduce_mean(
tf.cast(self.correct_prediction, "float"))
self.loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(
labels=self.labels, logits=self.train_digits))
self.train_op = optimizer.minimize(self.loss)
def testNoBatchNormScaleByDefault(self):
height, width = 224, 224
num_classes = 1000
inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope(inception.inception_v1_arg_scope()):
inception.inception_v1(inputs, num_classes, is_training=False)
self.assertEqual(tf.global_variables('.*/BatchNorm/gamma:0$'), [])
def testNoBatchNormScaleByDefault(self):
height, width = 224, 224
num_classes = 1000
inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope(inception.inception_v1_arg_scope()):
inception.inception_v1(inputs, num_classes, is_training=False)
self.assertEqual(tf.global_variables('.*/BatchNorm/gamma:0$'), [])
def testBatchNormScale(self):
height, width = 224, 224
num_classes = 1000
inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope(
inception.inception_v1_arg_scope(batch_norm_scale=True)):
inception.inception_v1(inputs, num_classes, is_training=False)
gamma_names = set(
v.op.name for v in tf.global_variables('.*/BatchNorm/gamma:0$'))
self.assertGreater(len(gamma_names), 0)
for v in tf.global_variables('.*/BatchNorm/moving_mean:0$'):
self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names)
def testBatchNormScale(self):
height, width = 224, 224
num_classes = 1000
inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope(
inception.inception_v1_arg_scope(batch_norm_scale=True)):
inception.inception_v1(inputs, num_classes, is_training=False)
gamma_names = set(
v.op.name for v in tf.global_variables('.*/BatchNorm/gamma:0$'))
self.assertGreater(len(gamma_names), 0)
for v in tf.global_variables('.*/BatchNorm/moving_mean:0$'):
self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names)
def testTrainEvalWithReuse(self):
train_batch_size = 5
eval_batch_size = 2
height, width = 224, 224
num_classes = 1000
train_inputs = tf.random_uniform((train_batch_size, height, width, 3))
inception.inception_v1(train_inputs, num_classes)
eval_inputs = tf.random_uniform((eval_batch_size, height, width, 3))
logits, _ = inception.inception_v1(eval_inputs, num_classes, reuse=True)
predictions = tf.argmax(logits, 1)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size,))
def __set_up_inception(self):
self.num_clss = self.config["Model"]["InceptionV1"]["num_clss"]
self.map_path = self.config["Model"]["InceptionV1"]["map_path"]
self.X = tf.placeholder(tf.float32,
shape=[
None, self.config["image"]["height"],
self.config["image"]["width"],
self.config["image"]["num_channels"]
])
with slim.arg_scope(inception.inception_v1_arg_scope()):
self.logits, self.end_points = inception.inception_v1(
self.X,
num_classes=self.num_clss,
is_training=False,
m_w=self.config["mantissa_width"],
e_w=self.config["exponent_width"],
s_w=self.config["shared_exponent_width"],
alter=self.config["quantize"],
debug=self.config["debug_layer"])
self.predictions = self.end_points["Predictions"]
self.variables_to_restore = slim.get_variables_to_restore()
self.init_assign_fn = slim.assign_from_checkpoint_fn(
self.config["to_ckpt"] +
self.config["Model"]["InceptionV1"]["ckpt_path"],
self.variables_to_restore)
def testTrainEvalWithReuse(self):
train_batch_size = 5
eval_batch_size = 2
height, width = 224, 224
num_classes = 1000
train_inputs = tf.random_uniform((train_batch_size, height, width, 3))
inception.inception_v1(train_inputs, num_classes)
eval_inputs = tf.random_uniform((eval_batch_size, height, width, 3))
logits, _ = inception.inception_v1(eval_inputs, num_classes, reuse=True)
predictions = tf.argmax(logits, 1)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size,))
def build(self, dataset, image_height, image_width, num_classes, is_training):
images, labels = self._load_batch(dataset, height=image_height, width=image_width)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception_v1_arg_scope()):
logits, end_points = inception_v1(images, num_classes=num_classes, is_training=is_training)
# Specify the loss function:
one_hot_labels = slim.one_hot_encoding(labels, num_classes)
loss = slim.losses.softmax_cross_entropy(logits, one_hot_labels)
# Create some summaries to visualize the training process:
tf.summary.scalar('losses/Total Loss', loss)
if is_training:
# Specify the optimizer and create the train op:
optimizer = tf.train.RMSPropOptimizer(learning_rate=0.01)
variables_to_train = []
for scope in self.trainable_scopes:
variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope)
variables_to_train.extend(variables)
variables = slim.get_model_variables("InceptionV1/Logits")
exec_op = slim.learning.create_train_op(loss, optimizer, variables_to_train=variables)
else:
exec_op = end_points['Predictions']
if self.checkpoints_file is not None:
self.init_fn = self._get_init_fn()
return exec_op, tf.get_default_graph()
def evaluate_inceptionV1_bin(batch_size):
with tf.Graph().as_default():
tf.logging.set_verbosity(tf.logging.INFO)
global_step = slim.get_or_create_global_step()
# Summary
dataset = plantVSnoplant.get_split('validation', plantVSnoplant_data_dir)
images,labels = load_batch(dataset, batch_size = batch_size, k=0, r=[], is_training =False)
with slim.arg_scope(inception.inception_v1_arg_scope(weight_decay=weight_decay)):
logits, _ = inception.inception_v1(images, num_classes=2, is_training=False)
total_output = []
total_labels = []
total_images = []
with tf.Session() as sess:
coord = tf.train.Coordinator()
saver = tf.train.Saver()
saver.restore(sess, tf.train.latest_checkpoint(train_inceptionV1_bin_dir))
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
for i in range(batch_size):
print('step: %d/%d' % (i, batch_size))
o, l , image= sess.run([logits, labels, images[0]])
o = tf.reduce_sum(o, 0)/float(5)
total_output.append(o)
total_labels.append(l[0])
total_images.append(image)
coord.request_stop()
coord.join(threads)
total_output = tf.stack(total_output,0)
total_output = tf.nn.softmax(total_output)
labels = tf.constant(total_labels)
total_images = sess.run(tf.stack(total_images,0))
top1_op = tf.nn.in_top_k(total_output, labels, 1)
top1_acc = sess.run(tf.reduce_mean(tf.cast(top1_op, tf.float32)))
print(top1_acc)
top5_op = tf.nn.in_top_k(total_output, labels, 5)
top5_acc = sess.run(tf.reduce_mean(tf.cast(top5_op, tf.float32)))
print(top5_acc)
accuracy1_sum = tf.summary.scalar('top1_accuracy', top1_acc)
accuracy5_sum = tf.summary.scalar('top5_accuracy', top5_acc)
images_sum = tf.summary.image("input_images", total_images, batch_size)
accuracy1, accuracy5, image_batch, step = sess.run([accuracy1_sum,accuracy5_sum,images_sum, global_step])
writer = tf.summary.FileWriter(eval_inceptionV1_bin_dir)
writer.add_summary(accuracy1, step)
writer.add_summary(accuracy5, step)
writer.add_summary(image_batch)
def testLogitsNotSqueezed(self):
num_classes = 25
images = tf.random_uniform([1, 224, 224, 3])
logits, _ = inception.inception_v1(images, num_classes=num_classes, spatial_squeeze=False)
with self.test_session() as sess:
tf.global_variables_initializer().run()
logits_out = sess.run(logits)
self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes])
def testLogitsNotSqueezed(self):
num_classes = 25
images = tf.random_uniform([1, 224, 224, 3])
logits, _ = inception.inception_v1(images, num_classes=num_classes, spatial_squeeze=False)
with self.test_session() as sess:
tf.initialize_all_variables().run()
logits_out = sess.run(logits)
self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes])
def train_inceptionV1(log_steps,save_summaries_sec,save_interval_secs,num_iterations = num_iterations_inception):
with tf.Graph().as_default():
tf.logging.set_verbosity(tf.logging.INFO)
summaries = set(tf.get_collection(tf.GraphKeys.SUMMARIES))
global_step = slim.get_or_create_global_step()
dataset = plantclef2015.get_split('train', plant_data_dir)
images,labels = load_batch(dataset, batch_size = batch_size, k=num_patches_inception, r=r_rotations_inception)
# Add Images to summaries
summaries.add(tf.summary.image("input_images", images, batch_size))
# Create the models
with slim.arg_scope(inception.inception_v1_arg_scope(weight_decay=weight_decay)):
logits, _ = inception.inception_v1(images, num_classes=1000, is_training=True)
# Specify the loss function:
one_hot_labels = slim.one_hot_encoding(labels, dataset.num_classes)
loss(logits, one_hot_labels)
#slim.losses.softmax_cross_entropy(logits, one_hot_labels)
total_loss = slim.losses.get_total_loss()
# Create some summaries to visualize the training process:
for variable in slim.get_model_variables():
summaries.add(tf.summary.histogram(variable.op.name, variable))
summaries.add(tf.summary.scalar('losses/Total_Loss', total_loss))
# Specify the optimizer and create the train op:
learning_rate = tf.train.exponential_decay(start_learning_rate, global_step,updating_iteration_for_learning_rate, updating_gamma, staircase=True)
optimizer = tf.train.MomentumOptimizer(learning_rate= learning_rate, momentum=momentum)
train_op = slim.learning.create_train_op(total_loss, optimizer)
summaries.add(tf.summary.scalar('training/Learning_Rate', learning_rate))
summary_op = tf.summary.merge(list(summaries), name='summary_op')
# Run the training:
final_loss = slim.learning.train(
train_op,
logdir=train_inceptionV1_dir,
log_every_n_steps=log_steps,
global_step=global_step,
number_of_steps= num_iterations,
summary_op=summary_op,
init_fn=get_init_fn_V1(),
save_summaries_secs=save_summaries_sec,
save_interval_secs=save_interval_secs)
print('Finished training. Last batch loss %f' % final_loss)
def extract_feat(frames_name):
"""
Reference:
1. Vasili's codes
2. https://github.com/tensorflow/models/issues/429#issuecomment-277885861
"""
slim_dir = "/home/xyang/workspace/models/research/slim"
checkpoints_dir = slim_dir + "/pretrain"
checkpoints_file = checkpoints_dir + '/inception_v1.ckpt'
batch_size = 256
sys.path.append(slim_dir)
from nets import inception
import tensorflow as tf
slim = tf.contrib.slim
image_size = inception.inception_v1.default_image_size
feat = []
with tf.Graph().as_default():
input_batch = tf.placeholder(dtype=tf.uint8,
shape=(batch_size, 300, 300, 3))
resized_images = tf.image.resize_images(
tf.image.convert_image_dtype(input_batch, dtype=tf.float32),
[image_size, image_size]
)
preprocessed_images = tf.multiply(tf.subtract(resized_images, 0.5), 2.0)
# Create the model, use the default arg scope to configure
# the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, endpoints = inception.inception_v1(preprocessed_images,
num_classes=1001,
is_training=False)
pool5 = endpoints['AvgPool_0a_7x7']
with tf.Session() as sess:
saver = tf.train.Saver()
saver.restore(sess, checkpoints_file)
for i in range(0, len(frames_name), batch_size):
print (i, '/', len(frames_name))
current_batch = np.zeros((batch_size, 300, 300, 3), dtype=np.uint8)
for j in range(batch_size):
if i+j == len(frames_name):
j -= 1
break
img = Image.open(frames_name[i+j]).convert('RGB').resize((300,300))
current_batch[j] = np.array(img)
feat_batch = sess.run(pool5, feed_dict={input_batch: current_batch})
feat_batch = np.squeeze(feat_batch)
feat.append(feat_batch[:j+1].astype('float32'))
return np.concatenate(feat, axis=0)
def predict(train_dir, test_dir, output_file, num_classes):
"""
Loads weights from a TF model and makes predictions.
Arguments:
train_dir: directory of trained model
test_dir: directory of test images (split into folders by class)
output_file: output file to store predictions
num_classes: number of classes of prediction
Returns:
Outpiuts a file output_file with predictions
"""
train_dir = train_dir
test_path = test_dir
output = output_file
nb_classes = num_classes
print('Start to read images!')
image_list = get_test_images(test_path)
processed_images = tf.placeholder(tf.float32, shape=(None, image_size, image_size, 3))
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(processed_images, num_classes=nb_classes, is_training=False)
def predict_fn(images):
return session.run(probabilities, feed_dict={processed_images: images})
probabilities = tf.nn.softmax(logits)
#Loads in latest training checkpoint
checkpoint_path = tf.train.latest_checkpoint(train_dir)
init_fn = slim.assign_from_checkpoint_fn(checkpoint_path,slim.get_variables_to_restore())
init_fn(session)
print('Start to transform images!')
images = transform_img_fn(image_list)
fto = open(output, 'w')
print('Start doing predictions!')
preds = predict_fn(images)
print (len(preds))
for p in range(len(preds)):
print (image_list[p], preds[p,:], np.argmax(preds[p,:]))
fto.write(image_list[p])
for j in range(len(preds[p,:])):
fto.write('\t' + str(preds[p, j]))
fto.write('\n')
fto.close()
def adapt_pretrain_imagenet_to_flower():
import os
from datasets import flowers
from nets import inception
from preprocessing import inception_preprocessing
slim = tf.contrib.slim
image_size = inception.inception_v1.default_image_size
def get_init_fn():
checkpoint_exclude_scopes = [
'InceptionV1/Logits', 'InceptionV1/AuxLogits'
]
exclusions = [scope.strip() for scope in checkpoint_exclude_scopes]
variables_to_restore = []
for var in slim.get_model_variables():
excluded = False
for exclusion in exclusions:
if var.op.name.startswith(exclusion):
excluded = True
break
if not excluded:
variables_to_restore.append(var)
return slim.assign_from_checkpoint_fn(
os.path.join(checkpoint_dir, 'inception_v1.ckpt'),
variables_to_restore)
with tf.Graph().as_default():
tf.logging.set_verbosity(tf.logging.INFO)
dataset = flowers.get_split('train', flower_data_dir)
images, _, labels = load_batch(dataset,
height=image_size,
width=image_size)
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(
images, num_classes=dataset.num_classes, is_training=True)
one_hot_labels = slim.one_hot_encoding(labels,
dataset.num_classes)
slim.losses.softmax_cross_entropy(logits, one_hot_labels)
total_loss = slim.losses.get_total_loss()
tf.scalar_summary('losses/Total Loss', total_loss)
optimizer = tf.train.AdamOptimizer(learning_rate=0.01)
train_op = slim.learning.create_train_op(total_loss, optimizer)
final_loss = slim.learning.train(train_op,
logdir=train_dir,
init_fn=get_init_fn(),
number_of_steps=2)
print("Finished training. Las batch loss %f" % final_loss)
def testBuildPreLogitsNetwork(self):
batch_size = 5
height, width = 224, 224
num_classes = None
inputs = tf.random_uniform((batch_size, height, width, 3))
net, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(net.op.name.startswith('InceptionV1/Logits/AvgPool'))
self.assertListEqual(net.get_shape().as_list(), [batch_size, 1, 1, 1024])
self.assertFalse('Logits' in end_points)
self.assertFalse('Predictions' in end_points)
def testBuildClassificationNetwork(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = tf.random_uniform((batch_size, height, width, 3))
logits, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith("InceptionV1/Logits"))
self.assertListEqual(logits.get_shape().as_list(), [batch_size, num_classes])
self.assertTrue("Predictions" in end_points)
self.assertListEqual(end_points["Predictions"].get_shape().as_list(), [batch_size, num_classes])
def testBuildClassificationNetwork(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = tf.random_uniform((batch_size, height, width, 3))
logits, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV1/Logits'))
self.assertListEqual(logits.get_shape().as_list(), [batch_size, num_classes])
self.assertTrue('Predictions' in end_points)
self.assertListEqual(end_points['Predictions'].get_shape().as_list(), [batch_size, num_classes])
def get_network_logits_and_endpoints(network, images):
if(network == 'inceptionV1'):
with slim.arg_scope(inception.inception_v1_arg_scope(weight_decay=weight_decay)):
logits, endpoints = inception.inception_v1(images, num_classes=1000, is_training=False)
elif(network == 'vgg16'):
with slim.arg_scope(vgg.vgg_arg_scope(weight_decay=weight_decay)):
logits, endpoints = vgg.vgg_16(images, num_classes=1000, is_training=False)
return logits,endpoints
def testBuildPreLogitsNetwork(self):
batch_size = 5
height, width = 224, 224
num_classes = None
inputs = tf.random_uniform((batch_size, height, width, 3))
net, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(net.op.name.startswith('InceptionV1/Logits/AvgPool'))
self.assertListEqual(net.get_shape().as_list(), [batch_size, 1, 1, 1024])
self.assertFalse('Logits' in end_points)
self.assertFalse('Predictions' in end_points)
def testEvaluation(self):
batch_size = 2
height, width = 224, 224
num_classes = 1000
eval_inputs = tf.random_uniform((batch_size, height, width, 3))
logits, _ = inception.inception_v1(eval_inputs, num_classes, is_training=False)
predictions = tf.argmax(logits, 1)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,))
def testEvaluation(self):
batch_size = 2
height, width = 224, 224
num_classes = 1000
eval_inputs = tf.random_uniform((batch_size, height, width, 3))
logits, _ = inception.inception_v1(eval_inputs, num_classes, is_training=False)
predictions = tf.argmax(logits, 1)
with self.test_session() as sess:
sess.run(tf.initialize_all_variables())
output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,))
def apply_pretrained_model():
import numpy as np
import os
import tensorflow as tf
import urllib2
from datasets import imagenet
from nets import inception
from preprocessing import inception_preprocessing
slim = tf.contrib.slim
batch_size = 3
image_size = inception.inception_v1.default_image_size
with tf.Graph().as_default():
url = 'https://upload.wikimedia.org/wikipedia/commons/7/70/EnglishCockerSpaniel_simon.jpg'
image_string = urllib2.urlopen(url).read()
image = tf.image.decode_jpeg(image_string, channels=3)
processed_image = inception_preprocessing.preprocess_image(
image, image_size, image_size, is_training=False)
processed_images = tf.expand_dims(processed_image, 0)
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(processed_images,
num_classes=1001,
is_training=False)
probabilities = tf.nn.softmax(logits)
init_fn = slim.assign_from_checkpoint_fn(
os.path.join(checkpoint_dir, 'inception_v1.ckpt'),
slim.get_model_variables('InceptionV1'))
with tf.Session() as sess:
init_fn(sess)
np_image, probabilities = sess.run([image, probabilities])
probabilities = probabilities[0, 0:]
sorted_inds = [
i[0] for i in sorted(enumerate(-probabilities),
key=lambda x: x[1])
]
plt.figure()
plt.imshow(np_image.astype(np.uint8))
plt.axis('off')
plt.show()
names = imagenet.create_readable_names_for_imagenet_labels()
for i in range(5):
index = sorted_inds[i]
print('Probabity %0.2f%% => [%s]' %
(probabilities[index], names[index]))
def eval(adv_imgs, labels, x_inputs, total_score, total_count):
image = (((adv_imgs + 1.0) * 0.5) * 255.0)
processed_imgs_inv1 = preprocess_for_model(image, 'inception_v1')
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits_inc_v1, end_points_inc_v1 = inception.inception_v1(
processed_imgs_inv1, num_classes=FLAGS.num_classes, is_training=False, scope='InceptionV1', reuse=True)
pred_inception = tf.argmax(end_points_inc_v1['Predictions'], 1)
# rescale pixle range from [-1, 1] to [0, 255] for resnet_v1 and vgg's input
processed_imgs_res_v1_50 = preprocess_for_model(image, 'resnet_v1_50')
with slim.arg_scope(resnet_v1.resnet_arg_scope()):
logits_res_v1_50, end_points_res_v1_50 = resnet_v1.resnet_v1_50(
processed_imgs_res_v1_50, num_classes=FLAGS.num_classes, is_training=False, scope='resnet_v1_50', reuse=True)
end_points_res_v1_50['logits'] = tf.squeeze(end_points_res_v1_50['resnet_v1_50/logits'], [1, 2])
end_points_res_v1_50['probs'] = tf.nn.softmax(end_points_res_v1_50['logits'])
pred_resnet = tf.argmax(end_points_res_v1_50['probs'], 1)
processed_imgs_inv3 = preprocess_for_model(image, 'inception_v3')
with slim.arg_scope(inception_v3.inception_v3_arg_scope()):
logits_res_inception_v3, end_points_inception_v3 = inception_v3.inception_v3(
processed_imgs_inv3, num_classes=FLAGS.num_classes, is_training=False, scope='InceptionV3', reuse=True)
pred_inception_v3 = tf.argmax(end_points_inception_v3['Predictions'], 1)
processed_imgs_inv_res = preprocess_for_model(image, 'inception_resnet_v2')
with slim.arg_scope(inception_resnet_v2.inception_resnet_v2_arg_scope()):
logits_res_inception_resnet, end_points_inception_resnet = inception_resnet_v2.inception_resnet_v2(
processed_imgs_inv_res, num_classes=FLAGS.num_classes, is_training=False, scope='InceptionResnetV2')
pred_ince_res = tf.argmax(end_points_inception_resnet['Predictions'], 1)
for i in range(adv_imgs.shape[0]):
def f1(total_score, total_count):
total_score = tf.add(total_score, 64)
return total_score, total_count
def f2(total_score, total_count):
adv = (((adv_imgs[i] + 1.0) * 0.5) * 255.0)
ori = (((x_inputs[i] + 1.0) * 0.5) * 255.0)
diff = tf.reshape(adv, [-1, 3]) - tf.reshape(ori, [-1, 3])
distance = tf.reduce_mean(tf.sqrt(tf.reduce_sum(tf.square(diff), axis=1)))
total_score = tf.add(total_score, distance)
total_count = tf.add(total_count, 1)
return total_score, total_count
total_score, total_count = tf.cond(tf.equal(pred_inception[i], labels[i]), lambda: f2(total_score, total_count), lambda: f1(total_score, total_count))
total_score, total_count = tf.cond(tf.equal(pred_resnet[i], labels[i]), lambda: f2(total_score, total_count), lambda: f1(total_score, total_count))
# total_score, total_count = tf.cond(tf.equal(pred_inception_v3[i], labels[i]), lambda: f2(total_score, total_count), lambda: f1(total_score, total_count))
total_score, total_count = tf.cond(tf.equal(pred_ince_res[i], labels[i]), lambda: f2(total_score, total_count), lambda: f1(total_score, total_count))
return total_score, total_count
def testUnknowBatchSize(self):
batch_size = 1
height, width = 224, 224
num_classes = 1000
inputs = tf.placeholder(tf.float32, (None, height, width, 3))
logits, _ = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith("InceptionV1/Logits"))
self.assertListEqual(logits.get_shape().as_list(), [None, num_classes])
images = tf.random_uniform((batch_size, height, width, 3))
with self.test_session() as sess:
sess.run(tf.initialize_all_variables())
output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes))
def testUnknowBatchSize(self):
batch_size = 1
height, width = 224, 224
num_classes = 1000
inputs = tf.placeholder(tf.float32, (None, height, width, 3))
logits, _ = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV1/Logits'))
self.assertListEqual(logits.get_shape().as_list(), [None, num_classes])
images = tf.random_uniform((batch_size, height, width, 3))
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes))
def inception_v1(inputs, is_training, opts):
with slim.arg_scope(inception.inception_v1_arg_scope(
weight_decay=opts.weight_decay,
use_batch_norm=opts.use_batch_norm,
batch_norm_decay=opts.batch_norm_decay,
batch_norm_epsilon=opts.batch_norm_epsilon,
activation_fn=tf.nn.relu)):
return inception.inception_v1(
inputs,
num_classes=opts.num_classes,
is_training=is_training,
dropout_keep_prob=opts.dropout_keep_prob,
prediction_fn=slim.softmax,
spatial_squeeze=opts.spatial_squeeze,
reuse=None,
global_pool=opts.global_pool)
def __call__(self, x_input):
"""Constructs model and return probabilities for given input."""
reuse = True if self.built else None
x_input = image_normalize(x_input, normalization_method[0])
x_input = tf.image.resize_images(x_input, [224, 224])
with slim.arg_scope(inception.inception_v1_arg_scope()):
_, end_points = inception.inception_v1(
x_input,
num_classes=self.num_classes,
is_training=False,
reuse=reuse)
self.built = True
output = end_points['Predictions']
# Strip off the extra reshape op at the output
probs = output.op.inputs[0]
return probs
def testUnknownImageShape(self):
tf.reset_default_graph()
batch_size = 2
height, width = 224, 224
num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess:
inputs = tf.placeholder(tf.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith("InceptionV1/Logits"))
self.assertListEqual(logits.get_shape().as_list(), [batch_size, num_classes])
pre_pool = end_points["Mixed_5c"]
feed_dict = {inputs: input_np}
tf.initialize_all_variables().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024])
def model_fn(features, labels, mode):
from nets import inception
slim = tf.contrib.slim
labels = tf.squeeze(labels, axis=1)
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, end_points = inception.inception_v1(features,
num_classes=2,
is_training=True)
if mode == tf.estimator.ModeKeys.TRAIN:
loss = tf.reduce_mean(
tf.losses.sparse_softmax_cross_entropy(logits=logits,
labels=labels))
return TFEstimatorSpec(mode, predictions=logits, loss=loss)
else:
raise NotImplementedError
def testUnknownImageShape(self):
tf.reset_default_graph()
batch_size = 2
height, width = 224, 224
num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess:
inputs = tf.placeholder(tf.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV1/Logits'))
self.assertListEqual(logits.get_shape().as_list(), [batch_size, num_classes])
pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np}
tf.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024])
def eval_adapt_pretrained_to_flower():
import numpy as np
import tensorflow as tf
from datasets import flowers
from nets import inception
slim = tf.contrib.slim
image_size = inception.inception_v1.default_image_size
batch_size = 8
with tf.Graph().as_default():
tf.logging.set_verbosity(tf.logging.INFO)
dataset = flowers.get_split('train', flower_data_dir)
images, images_raw, labels = load_batch(dataset,
height=image_size,
width=image_size)
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(
images, num_classes=dataset.num_classes, is_training=True)
probabilities = tf.nn.softmax(logits)
checkpoint_path = tf.train.latest_checkpoint(train_dir)
init_fn = slim.assign_from_checkpoint_fn(
checkpoint_path, slim.get_variables_to_restore())
with tf.Session() as sess:
with slim.queues.QueueRunners(sess):
sess.run(tf.initialize_local_variables())
init_fn(sess)
np_probabilities, np_images_raw, np_labels = sess.run(
[probabilities, images_raw, labels])
for i in xrange(batch_size):
image = np_images_raw[i, :, :, :]
true_label = np_labels[i]
predicted_label = np.argmax(np_probabilities[i, :])
predicted_name = dataset.labels_to_names[
predicted_label]
true_name = dataset.labels_to_names[true_label]
plt.figure()
plt.imshow(image.astype(np.uint8))
plt.title('Ground Truth: [%s], Prediction [%s]' %
(true_name, predicted_name))
plt.axis('off')
plt.show()
def inception_tst():
with tf.Graph().as_default():
url = 'https://upload.wikimedia.org/wikipedia/commons/7/70/EnglishCockerSpaniel_simon.jpg'
image_string = urllib.urlopen(url).read()
image = tf.image.decode_jpeg(image_string, channels=3)
processed_image = inception_preprocessing.preprocess_image(
image, image_size, image_size, is_training=False)
processed_images = tf.expand_dims(processed_image, 0)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(processed_images,
num_classes=1001,
is_training=False)
probabilities = tf.nn.softmax(logits)
variables_to_restore = slim.get_variables()
print(len(variables_to_restore))
# pprint(variables_to_restore)
variables_to_restore = slim.get_model_variables()
print(len(variables_to_restore))
return
init_fn = slim.assign_from_checkpoint_fn(
os.path.join(checkpoints_dir, 'inception_v1.ckpt'),
slim.get_model_variables('InceptionV1'))
with tf.Session() as sess:
init_fn(sess)
np_image, probabilities = sess.run([image, probabilities])
probabilities = probabilities[0, 0:]
sorted_inds = [
i[0]
for i in sorted(enumerate(-probabilities), key=lambda x: x[1])
]
plt.figure()
plt.imshow(np_image.astype(np.uint8))
plt.axis('off')
plt.show()
names = imagenet.create_readable_names_for_imagenet_labels()
for i in range(5):
index = sorted_inds[i]
print('Probability %0.2f%% => [%s]' %
(probabilities[index] * 100, names[index]))
def predict(train_dir, test_dir, num_classes):
"""
Loads weights from a TF model and makes predictions.
Arguments:
train_dir: directory of trained model
test_dir: directory of test images (split into folders by class)
num_classes: number of classes of prediction
Returns:
Returns logits in the order of test images given and nparray of predictions
"""
processed_images = tf.placeholder(tf.float32,
shape=(None, image_size, image_size, 3))
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(processed_images,
num_classes=num_classes,
is_training=False)
probabilities = tf.nn.softmax(logits)
def predict_fn(images):
return session.run(probabilities, feed_dict={processed_images: images})
#Loads in latest training checkpoint
checkpoint_path = tf.train.latest_checkpoint(train_dir)
init_fn = slim.assign_from_checkpoint_fn(checkpoint_path,
slim.get_variables_to_restore())
init_fn(session)
image_list = get_test_images(test_dir)
images = transform_img_fn(image_list)
predicted_probs = predict_fn(images)
predictions = np.empty((len(predicted_probs), ))
for example in range(len(predicted_probs)):
predictions[example] = np.argmax(predicted_probs[example, :])
return predicted_probs, predictions
def label_test():
image_size = inception.inception_v1.default_image_size
batch_size = 12
with tf.Graph().as_default():
tf.logging.set_verbosity(tf.logging.INFO)
dataset = flowers.get_split('validation', flowers_data_dir)
images, images_raw, labels = load_batch(dataset,
height=image_size,
width=image_size)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(images,
num_classes=dataset.num_classes,
is_training=True)
probabilities = tf.nn.softmax(logits)
checkpoint_path = tf.train.latest_checkpoint(train_dir)
init_fn = slim.assign_from_checkpoint_fn(
checkpoint_path, slim.get_variables_to_restore())
with tf.Session() as sess:
with slim.queues.QueueRunners(sess):
sess.run(tf.initialize_local_variables())
init_fn(sess)
np_probabilities, np_images_raw, np_labels = sess.run(
[probabilities, images, labels])
for i in range(batch_size):
image = np_images_raw[i, :, :, :]
true_label = np_labels[i]
predicted_label = np.argmax(np_probabilities[i, :])
predicted_name = dataset.labels_to_names[predicted_label]
true_name = dataset.labels_to_names[true_label]
plt.figure()
plt.imshow(image / (image.max() - image.min()))
# plt.imshow(image.astype(np.uint8))
plt.title('Ground Truth: [%s], Prediction [%s]' %
(true_name, predicted_name))
plt.axis('off')
plt.show()
def load_inception_model(network_config, net_input=None):
with slim.arg_scope(inception.inception_v1_arg_scope()):
if net_input is None:
net_input = tf.placeholder("float",
[None, image_size, image_size, 3],
name='nn_input')
logits, end_points = inception.inception_v1(net_input,
num_classes=1001,
is_training=True)
init_fn = slim.assign_from_checkpoint_fn(
os.path.join('./checkpoints', 'inception_v1.ckpt'),
slim.get_model_variables('InceptionV1'))
return logits, end_points, init_fn, net_input
def eval():
# This might take a few minutes.
image_size = inception.inception_v1.default_image_size
with tf.Graph().as_default():
#tf.logging.set_verbosity(tf.logging.INFO)
dataset = flowers.get_split('train', flowers_data_dir)
images, images_raw, labels = load_batch(dataset,
height=image_size,
width=image_size)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(images,
num_classes=dataset.num_classes,
is_training=True)
predictions = tf.argmax(logits, 1)
checkpoint_path = tf.train.latest_checkpoint(train_dir)
init_fn = slim.assign_from_checkpoint_fn(
checkpoint_path, slim.get_variables_to_restore())
# Define the metrics:
names_to_values, names_to_updates = slim.metrics.aggregate_metric_map({
'eval/Accuracy':
slim.metrics.streaming_accuracy(predictions, labels),
'eval/Recall@5':
slim.metrics.streaming_sparse_recall_at_k(logits, labels, 5)
})
print('Running evaluation Loop...')
checkpoint_path = tf.train.latest_checkpoint(train_dir)
metric_values = slim.evaluation.evaluate_once(
master='',
checkpoint_path=checkpoint_path,
logdir=train_dir,
eval_op=list(names_to_updates.values()),
final_op=list(names_to_values.values()))
names_to_values = dict(zip(names_to_values.keys(), metric_values))
for name in names_to_values:
print('%s: %f' % (name, names_to_values[name]))
def build(self, dataset, image_height, image_width, num_classes,
is_training):
images, labels = self._load_batch(dataset,
height=image_height,
width=image_width)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception_v1_arg_scope()):
logits, end_points = inception_v1(images,
num_classes=num_classes,
is_training=is_training)
# Specify the loss function:
one_hot_labels = slim.one_hot_encoding(labels, num_classes)
loss = slim.losses.softmax_cross_entropy(logits, one_hot_labels)
# Create some summaries to visualize the training process:
tf.summary.scalar('losses/Total Loss', loss)
if is_training:
# Specify the optimizer and create the train op:
optimizer = tf.train.RMSPropOptimizer(learning_rate=0.01)
variables_to_train = []
for scope in self.trainable_scopes:
variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,
scope)
variables_to_train.extend(variables)
variables = slim.get_model_variables("InceptionV1/Logits")
exec_op = slim.learning.create_train_op(
loss, optimizer, variables_to_train=variables)
else:
exec_op = end_points['Predictions']
if self.checkpoints_file is not None:
self.init_fn = self._get_init_fn()
return exec_op, tf.get_default_graph()
def fine_tune():
with tf.Graph().as_default():
tf.logging.set_verbosity(tf.logging.INFO)
dataset = flowers.get_split('train', flowers_data_dir)
images, _, labels = load_batch(dataset,
height=image_size,
width=image_size)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(images,
num_classes=dataset.num_classes,
is_training=True)
# Specify the loss function:
one_hot_labels = slim.one_hot_encoding(labels, dataset.num_classes)
slim.losses.softmax_cross_entropy(logits, one_hot_labels)
total_loss = slim.losses.get_total_loss()
# Create some summaries to visualize the training process:
tf.summary.scalar('losses/Total Loss', total_loss)
# Specify the optimizer and create the train op:
optimizer = tf.train.AdamOptimizer(learning_rate=0.01)
train_op = slim.learning.create_train_op(total_loss, optimizer)
# Run the training:
final_loss = slim.learning.train(
train_op,
logdir=train_dir,
init_fn=get_init_fn(),
number_of_steps=1000, # For speed, we just do 1 epoch
save_interval_secs=600,
save_summaries_secs=6000,
log_every_n_steps=1,
)
print('Finished training. Last batch loss %f' % final_loss)
EPOCHS = 1
classes = util.pkl_load(workspace.class_pkl)
csv_files = workspace.test_csvs
graph = tf.Graph()
with graph.as_default():
tf.logging.set_verbosity(tf.logging.INFO)
fp_placeholder = tf.placeholder(tf.string)
image = read_one_image(fp_placeholder)
images = tf.reshape(image, [1, 224, 224, 3])
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(images, num_classes=len(classes), is_training=False)
probabilities = tf.nn.softmax(logits)
checkpoint = tf.train.latest_checkpoint(os.path.join(workspace.inception_cpkt, 'log'))
init_fn = slim.assign_from_checkpoint_fn(
checkpoint,
slim.get_model_variables())
with tf.Session(graph=graph) as sess:
init_fn(sess)
top1 = 0
top5 = 0
total = 0
# there is prob a better way to do this will all TF but
# this will do for calculating accuracy.
for csv_fp in workspace.test_csvs:
slim = tf.contrib.slim
image_size = inception.inception_v1.default_image_size
with tf.Graph().as_default():
#url = 'https://upload.wikimedia.org/wikipedia/commons/7/70/EnglishCockerSpaniel_simon.jpg'
#url = 'https://pbs.twimg.com/media/B5B9ZsvIgAATiC6.jpg'
url=sys.argv[1]
image_string = urllib2.urlopen(url).read()
image = tf.image.decode_jpeg(image_string, channels=3)
processed_image = inception_preprocessing.preprocess_image(image, image_size, image_size, is_training=False)
processed_images = tf.expand_dims(processed_image, 0)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(inception.inception_v1_arg_scope()):
logits, _ = inception.inception_v1(processed_images, num_classes=1001, is_training=False)
probabilities = tf.nn.softmax(logits)
checkpoints_dir='slim_pretrained'
init_fn = slim.assign_from_checkpoint_fn(
os.path.join(checkpoints_dir, 'inception_v1.ckpt'),
slim.get_variables_to_restore())
#slim.get_model_variables())
#slim.get_model_variables('InceptionV1'))
#saver= tf.train.Saver(tf.all_variables())
#saver= tf.train.Saver(list(set(slim.get_model_variables()) | set(tf.trainable_variables())))
with tf.Session() as sess:
init_fn(sess)
np_image, probabilities = sess.run([image, probabilities])
probabilities = probabilities[0, 0:]
sorted_inds = [i[0] for i in sorted(enumerate(-probabilities), key=lambda x:x[1])]
#saver.save(sess,'/home/sankit/Documents/cv-tricks/models/slim/my-saved/inception_v1.ckpt')