def testAddEvaluationStep(self):
with tf.Graph().as_default():
with tf.Session() as sess:
tf.placeholder(tf.float32, [1], name='final')
tf.placeholder(tf.float32, [1], name='gt')
self.assertIsNotNone(
retrain.add_evaluation_step(sess.graph, 'final', 'gt'))
def evaluate_graph(graph_file_name):
with load_graph(graph_file_name).as_default() as graph:
ground_truth_input = tf.placeholder(tf.float32, [None, 5],
name='GroundTruthInput')
image_buffer_input = graph.get_tensor_by_name('input:0')
final_tensor = graph.get_tensor_by_name('final_result:0')
accuracy, _ = retrain.add_evaluation_step(final_tensor,
ground_truth_input)
logits = graph.get_tensor_by_name("final_training_ops/Wx_plus_b/add:0")
xent = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(labels=ground_truth_input,
logits=logits))
image_dir = 'tf_files/flower_photos'
testing_percentage = 10
validation_percentage = 10
validation_batch_size = 100
category = 'testing'
image_lists = retrain.create_image_lists(image_dir, testing_percentage,
validation_percentage)
class_count = len(image_lists.keys())
ground_truths = []
filenames = []
for label_index, label_name in enumerate(image_lists.keys()):
for image_index, image_name in enumerate(
image_lists[label_name][category]):
image_name = retrain.get_image_path(image_lists, label_name,
image_index, image_dir,
category)
ground_truth = np.zeros([1, class_count], dtype=np.float32)
ground_truth[0, label_index] = 1.0
ground_truths.append(ground_truth)
filenames.append(image_name)
accuracies = []
xents = []
with tf.Session(graph=graph) as sess:
for filename, ground_truth in zip(filenames, ground_truths):
image = Image.open(filename).resize((224, 224), Image.ANTIALIAS)
image = np.array(image, dtype=np.float32)[None, ...]
image = (image - 128) / 128.0
feed_dict = {
image_buffer_input: image,
ground_truth_input: ground_truth
}
eval_accuracy, eval_xent = sess.run([accuracy, xent], feed_dict)
accuracies.append(eval_accuracy)
xents.append(eval_xent)
return np.mean(accuracies), np.mean(xents)
def testAddEvaluationStep(self):
with tf.Graph().as_default():
final = tf.placeholder(tf.float32, [1], name='final')
gt = tf.placeholder(tf.int64, [1], name='gt')
self.assertIsNotNone(retrain.add_evaluation_step(final, gt))
workbook = xlsxwriter.Workbook("ensemble_accuracy.xlsx")
worksheet = workbook.add_worksheet()
with graphs.as_default() as gs:
dataset = dataset_factory.get_dataset(FLAGS.dataset_name,
FLAGS.dataset_split_name,
FLAGS.dataset_dir)
ground_truth_input = tf.placeholder(tf.float32,
[None, dataset.num_classes],
name='GroundTruthInput')
predictions_average = tf.placeholder(tf.float32,
[None, dataset.num_classes],
name='predictions_average')
accuracy_eval, _ = retrain.add_evaluation_step(predictions_average,
ground_truth_input)
logger.info('')
if pre_ensemble_num < 50:
pre_ensemble_num = 50
random_times = 100
ensemble_num = int(pre_ensemble_num * 0.2)
logger.info("random ensemble num is : %d , random time is: %d" %
(ensemble_num, random_times))
#sameple will can do the shuffle func
if FLAGS.base_mode == True:
pre_ensemble_df = all_files_exclude_base_df[0:ensemble_num]
ensemble_df = pre_ensemble_df[:].sample(n=ensemble_num)
def extract():
#sess=tf.Session()
#先加载图和参数变量
# for op in graph.get_operations():
# print(str(op.name))
# var = tf.global_variables()#全部调用
# for i in var:
# print(i)
input_layer = "input"
#nput_layer = "MobilenetV2/input"
output_layer = "MobilenetV2/Predictions/Reshape_1"
#output_layer= "MobilenetV2/Logits/output"
#graph = load_graph('./frozen_pb/frozen_0-0-0.pb')
#with graph.as_default() as g:
# image_buffer_input = g.get_tensor_by_name('input:0')
# final_tensor = g.get_tensor_by_name('MobilenetV2/Logits/AvgPool:0')
#image_dir = '/home/xxxx/PHICOMM/ai-share/dataset/imagenet/raw-data/imagenet-data/validation'
image_dir = "/home/deepl/PHICOMM/dataset/cifar10_tf/cifar-10/test"
testing_percentage = 100
validation_percentage = 0
category = 'testing'
image_lists = retrain.create_image_lists(image_dir, testing_percentage,
validation_percentage)
class_count = len(image_lists.keys())
print(class_count)
total_start = time.time()
ground_truths = []
filenames = []
all_files_df = pd.DataFrame(
columns=['image_name', 'ground_truth', "predecit_label"])
for label_index, label_name in enumerate(image_lists.keys()):
for image_index, image_name in enumerate(
image_lists[label_name][category]):
image_name = retrain.get_image_path(image_lists, label_name,
image_index, image_dir,
category)
ground_truth = np.zeros([1, class_count], dtype=np.float32)
ground_truth[0, label_index] = 1.0
ground_truths.append(ground_truth)
filenames.append(image_name)
ground_truth_argmax = np.argmax(ground_truth, axis=1)
ground_truth_argmax = np.squeeze(ground_truth_argmax)
all_files_df = all_files_df.append(
[{
'image_name': image_name,
'ground_truth': ground_truth_argmax
}],
ignore_index=True)
#all_files_df.to_csv("ground_truth1.csv")
#print(filenames)
if os.path.exists("./data"):
print("data is exist, please delete it!")
exit()
#shutil.rmtree("./data")
#os.makedirs("./data")
#sio.savemat('./data/truth.mat',{"truth": ground_truths})
cf = 0.875
predictions = []
i = 0
start = time.time()
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
#with tf.Session(graph=graph) as sess:
network_fn = nets_factory.get_network_fn("mobilenet_v2",
num_classes=10,
is_training=False)
image_size = 224
placeholder = tf.placeholder(name='input',
dtype=tf.float32,
shape=[128, image_size, image_size, 3])
logits, _ = network_fn(placeholder)
graph = tf.get_default_graph()
saver = tf.train.Saver()
#raph_def = graph.as_graph_def()
#aver = tf.train.import_meta_graph(graph_def)
#saver = tf.train.import_meta_graph('./mobilenetv2_on_cifar10_check_point/0/model_0/model.ckpt-20000.meta')
saver.restore(
sess,
'./mobilenetv2_on_cifar10_check_point/0/model_0/model.ckpt-20000')
# Initalize the variables
#sess.run(tf.global_variables_initializer())
output_operation = graph.get_operation_by_name(output_layer)
input_operation = graph.get_operation_by_name(input_layer)
read_tensor_from_jpg_image_file_op = read_tensor_from_jpg_image_file(
input_height=224, input_width=224)
file_num = len(filenames)
batch_size = 128
count = file_num // batch_size
print("need %d batch, every batch is %d" % (count, batch_size))
for i in range(count):
print("this is %d batch" % i)
print("jpg order get from %d to %d" %
(batch_size * i, batch_size * i + batch_size - 1))
for j in range(batch_size):
if j == 0:
t_batch = sess.run(read_tensor_from_jpg_image_file_op,
feed_dict={
"fnamejpg:0":
filenames[batch_size * i + j]
})
t_batch = np.expand_dims(t_batch, axis=0)
else:
t = sess.run(read_tensor_from_jpg_image_file_op,
feed_dict={
"fnamejpg:0":
filenames[batch_size * i + j]
})
t = np.expand_dims(t, axis=0)
t_batch = np.concatenate((t_batch, t), axis=0)
print(t_batch.shape)
# feed_dict={image_buffer_input: t}
# ft = final_tensor.eval(feed_dict, sess)
# pre = sess.run(output_operation.outputs[0],
# {input_operation.outputs[0]: t_batch})
pre = sess.run(logits, {input_operation.outputs[0]: t_batch})
#print(pre.shape)
for k in range(batch_size):
predictions.append(pre[k, :])
if file_num % batch_size != 0:
print("this is %d batch" % count)
print("jpg order get from %d to %d" %
(batch_size * count, file_num - 1))
extra_num = file_num - count * batch_size
for j in range(extra_num):
if j == 0:
t_batch = sess.run(read_tensor_from_jpg_image_file_op,
feed_dict={
"fnamejpg:0":
filenames[batch_size * count + j]
})
t_batch = np.expand_dims(t_batch, axis=0)
else:
t = sess.run(read_tensor_from_jpg_image_file_op,
feed_dict={
"fnamejpg:0":
filenames[batch_size * count + j]
})
t = np.expand_dims(t, axis=0)
t_batch = np.concatenate((t_batch, t), axis=0)
print(t_batch.shape)
for j in range(extra_num, batch_size):
t_batch = np.concatenate((t_batch, t), axis=0)
print(t_batch.shape)
# pre = sess.run(output_operation.outputs[0],
# {input_operation.outputs[0]: t_batch})
pre = sess.run(logits, {input_operation.outputs[0]: t_batch})
for k in range(extra_num):
predictions.append(pre[k, :])
#i = i + 1
#print(i)
predictions = np.array(predictions)
#print(predictions.shape)
#predictions = np.squeeze(predictions)
ground_truths = np.array(ground_truths)
print(ground_truths.shape)
ground_truths = np.squeeze(ground_truths)
print(predictions.shape)
print(ground_truths.shape)
with tf.Session(config=config) as sess:
# with tf.Session(graph=graph) as sess:
ground_truth_input = tf.placeholder(tf.float32, [None, 10],
name='GroundTruthInput')
fts = tf.placeholder(tf.float32, [None, 10], name='fts')
accuracy, _ = retrain.add_evaluation_step(fts, ground_truth_input)
feed_dict = {fts: predictions, ground_truth_input: ground_truths}
#accuracies.append(accuracy.eval(feed_dict, sess))
ret = accuracy.eval(feed_dict, sess)
for index, row in all_files_df.iterrows():
row['predecit_label'] = np.squeeze(
np.argmax(predictions[index, :], axis=0))
all_files_df.to_csv("ground_truth.csv")
print('Ensemble Accuracy: %g' % ret)
stop = time.time()
#print(str((stop-start)/len(ftg))+' seconds.')
#sio.savemat('./data/feature.mat',{"feature": ftg})
total_stop = time.time()
print("total time is " + str((total_stop - total_start)) + ' seconds.')
def extract():
input_layer = "input"
output_layer = "MobilenetV2/Predictions/Reshape_1"
graph = load_graph('./frozen_pb/frozen_0-0-0.pb')
#with graph.as_default() as g:
# image_buffer_input = g.get_tensor_by_name('input:0')
# final_tensor = g.get_tensor_by_name('MobilenetV2/Logits/AvgPool:0')
input_operation = graph.get_operation_by_name(input_layer)
output_operation = graph.get_operation_by_name(output_layer)
#image_dir = '/home/xxxx/PHICOMM/ai-share/dataset/imagenet/raw-data/imagenet-data/validation'
image_dir = "/home/deepl/PHICOMM/dataset/cifar10_tf/cifar-10/test"
testing_percentage = 100
validation_percentage = 0
category = 'testing'
image_lists = retrain.create_image_lists(image_dir, testing_percentage,
validation_percentage)
class_count = len(image_lists.keys())
print(class_count)
total_start = time.time()
ground_truths = []
filenames = []
all_files_df = pd.DataFrame(
columns=['image_name', 'ground_truth', "predecit_label"])
for label_index, label_name in enumerate(image_lists.keys()):
for image_index, image_name in enumerate(
image_lists[label_name][category]):
image_name = retrain.get_image_path(image_lists, label_name,
image_index, image_dir,
category)
ground_truth = np.zeros([1, class_count], dtype=np.float32)
ground_truth[0, label_index] = 1.0
ground_truths.append(ground_truth)
filenames.append(image_name)
ground_truth_argmax = np.argmax(ground_truth, axis=1)
ground_truth_argmax = np.squeeze(ground_truth_argmax)
all_files_df = all_files_df.append(
[{
'image_name': image_name,
'ground_truth': ground_truth_argmax
}],
ignore_index=True)
#all_files_df.to_csv("ground_truth.csv")
if os.path.exists("./data"):
print("data is exist, please delete it!")
exit()
#shutil.rmtree("./data")
#os.makedirs("./data")
#sio.savemat('./data/truth.mat',{"truth": ground_truths})
cf = 0.875
predictions = []
i = 0
start = time.time()
with tf.Session(graph=graph) as sess:
read_tensor_from_jpg_image_file_op = read_tensor_from_jpg_image_file(
input_height=224, input_width=224)
for filename in filenames:
t = sess.run(read_tensor_from_jpg_image_file_op,
feed_dict={"fnamejpg:0": filename})
t = np.expand_dims(t, axis=0)
#print(t.shape)
# feed_dict={image_buffer_input: t}
# ft = final_tensor.eval(feed_dict, sess)
pre = sess.run(output_operation.outputs[0],
{input_operation.outputs[0]: t})
predictions.append(pre)
#i = i + 1
#print(i)
predictions = np.array(predictions)
predictions = np.squeeze(predictions)
ground_truths = np.array(ground_truths)
ground_truths = np.squeeze(ground_truths)
print(predictions.shape)
print(ground_truths.shape)
with tf.Session(graph=graph) as sess:
ground_truth_input = tf.placeholder(tf.float32, [None, 10],
name='GroundTruthInput')
fts = tf.placeholder(tf.float32, [None, 10], name='fts')
accuracy, _ = retrain.add_evaluation_step(fts, ground_truth_input)
feed_dict = {fts: predictions, ground_truth_input: ground_truths}
#accuracies.append(accuracy.eval(feed_dict, sess))
ret = accuracy.eval(feed_dict, sess)
for index, row in all_files_df.iterrows():
row['predecit_label'] = np.squeeze(
np.argmax(predictions[index, :], axis=0))
all_files_df.to_csv("ground_truth.csv")
print('Ensemble Accuracy: %g' % ret)
stop = time.time()
#print(str((stop-start)/len(ftg))+' seconds.')
#sio.savemat('./data/feature.mat',{"feature": ftg})
total_stop = time.time()
print("total time is " + str((total_stop - total_start)) + ' seconds.')
num_classes=(dataset.num_classes - FLAGS.labels_offset),
is_training=False)
placeholder = tf.placeholder(name='input', dtype=tf.float32,
shape=[None, FLAGS.eval_image_size,
FLAGS.eval_image_size, 3])
logits, _ = network_fn(placeholder)
graph = tf.get_default_graph()
saver = tf.train.Saver()
output_operation = graph.get_operation_by_name(FLAGS.output_layer);
input_operation = graph.get_operation_by_name(FLAGS.input_layer);
ground_truth_input = tf.placeholder(
tf.float32, [None, dataset.num_classes], name='GroundTruthInput')
predicts = tf.placeholder(tf.float32, [None, dataset.num_classes], name='predicts')
accuracy, _ = retrain.add_evaluation_step(predicts, ground_truth_input)
index = 0
for i,checkpoint in enumerate(all_checkpoints):
checkpoint_prefix = checkpoint.replace('.data-00000-of-00001', '')
saver.restore(sess,checkpoint_prefix)
predictions = sess.run(output_operation.outputs[0],
{input_operation.outputs[0]: ftg})
#print(predictions.shape)
feed_dict={predicts: predictions, ground_truth_input: ground_truths}
ret = accuracy.eval(feed_dict, sess)
def extract():
input_layer= "input"
#nput_layer = "MobilenetV2/input"
output_layer= "MobilenetV2/Predictions/Reshape_1"
total_start = time.time()
# if os.path.exists("./data"):
# print("data is exist, please delete it!")
# exit()
#shutil.rmtree("./data")
#os.makedirs("./data")
#sio.savemat('./data/truth.mat',{"truth": ground_truths})
cf = 0.875
predictions = []
ground_truths = []
i = 0
start = time.time()
config = tf.ConfigProto()
config.gpu_options.allow_growth=True
with tf.Session(config=config) as sess:
dataset = dataset_factory.get_dataset(
FLAGS.dataset_name, FLAGS.dataset_split_name, FLAGS.dataset_dir)
network_fn = nets_factory.get_network_fn(
FLAGS.model_name,
num_classes=(dataset.num_classes - FLAGS.labels_offset),
is_training=False)
provider = slim.dataset_data_provider.DatasetDataProvider(
dataset,
num_epochs=1,
shuffle=False,
common_queue_capacity=2 * FLAGS.batch_size,
common_queue_min=FLAGS.batch_size)
[image, label] = provider.get(['image', 'label'])
label -= FLAGS.labels_offset
preprocessing_name = FLAGS.preprocessing_name or FLAGS.model_name
image_preprocessing_fn = preprocessing_factory.get_preprocessing(
preprocessing_name,
is_training=False)
eval_image_size = FLAGS.eval_image_size or network_fn.default_image_size
image = image_preprocessing_fn(image, eval_image_size, FLAGS.eval_image_size)
# filename_queue = tf.train.string_input_producer(
# [FLAGS.dataset_dir], num_epochs=1)
# image, label = read_and_decode(filename_queue,FLAGS.eval_image_size)
# print(image.shape)
# print(image)
# print(label.shape)
images_batch, labels_batch = tf.train.batch(
[image, label],
batch_size=FLAGS.batch_size,
num_threads=FLAGS.num_preprocessing_threads,
capacity=5 * FLAGS.batch_size,
allow_smaller_final_batch=True)
labels_batch_one_hot = tf.one_hot(labels_batch,dataset.num_classes)
placeholder = tf.placeholder(name='input', dtype=tf.float32,
shape=[None, FLAGS.eval_image_size,
FLAGS.eval_image_size, 3])
logits, _ = network_fn(placeholder)
graph = tf.get_default_graph()
saver = tf.train.Saver()
output_operation = graph.get_operation_by_name(output_layer);
input_operation = graph.get_operation_by_name(input_layer);
batch_size = FLAGS.batch_size
print("every batch is %d"%(batch_size))
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
saver.restore(sess,FLAGS.checkpoint_path)
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess, coord)
count = 0
try:
while not coord.should_stop():
image_batch_v, label_batch_v = sess.run([images_batch, labels_batch_one_hot])
#print(image_batch_v.shape, label_batch_v.shape)
print("this is %d batch"%count)
ground_truths.extend(label_batch_v)
count += 1
pre = sess.run(logits,
{input_operation.outputs[0]: image_batch_v})
predictions.extend(pre)
except tf.errors.OutOfRangeError:
print("done")
finally:
coord.request_stop()
coord.join(threads)
#i = i + 1
#print(i)
predictions = np.array(predictions)
ground_truths = np.array(ground_truths)
print(predictions.shape)
print(ground_truths.shape)
with tf.Session(config=config) as sess:
# with tf.Session(graph=graph) as sess:
ground_truth_input = tf.placeholder(
tf.float32, [None, 1001], name='GroundTruthInput')
predicts = tf.placeholder(tf.float32, [None, 1001], name='predicts')
accuracy, _ = retrain.add_evaluation_step(predicts, ground_truth_input)
feed_dict={predicts: predictions, ground_truth_input: ground_truths}
#accuracies.append(accuracy.eval(feed_dict, sess))
ret = accuracy.eval(feed_dict, sess)
print('Ensemble Accuracy: %g' % ret)
stop = time.time()
#print(str((stop-start)/len(ftg))+' seconds.')
#sio.savemat('./data/feature.mat',{"feature": ftg})
total_stop = time.time()
print("total time is "+str((total_stop-total_start))+' seconds.')
