def optimize(self):
utils.create_log_dir(self.hp_utils.save_path)
if not self.restart:
utils.check_conflict(self.hp_utils.save_path, check=self.check)
utils.create_log_dir(self.hp_utils.save_path)
dy = {y_name: None for y_name in self.hp_utils.y_names}
self.hp_utils.save_hp_conf(list(range(len(self.hp_utils.var_names))),
dy,
None,
record=False)
self.n_jobs = self.get_n_jobs()
self.check_double_submission()
save_time = utils.save_elapsed_time(self.hp_utils.save_path,
self.hp_utils.lock, self.verbose,
self.print_freq)
if self.n_parallels <= 1:
self._optimize_sequential(save_time)
else:
self._optimize_parallel(save_time)
hps_conf, losses = self.hp_utils.load_hps_conf(do_sort=True)
best_hp_conf, best_performance = hps_conf[0], losses[0][0]
self.print_optimized_result(best_hp_conf, best_performance)
return best_hp_conf, best_performance
def optimize(self):
utils.create_log_dir(self.hp_utils.save_path)
if not self.restart:
utils.check_conflict(self.hp_utils.save_path)
utils.create_log_dir(self.hp_utils.save_path)
self.n_jobs = self.get_n_jobs()
if self.n_parallels <= 1:
self._optimize_sequential()
else:
self._optimize_parallel()
hps_conf, losses = self.hp_utils.load_hps_conf(do_sort=True)
best_hp_conf, best_performance = hps_conf[0], losses[0]
self.print_optimized_result(best_hp_conf, best_performance[0])
return best_hp_conf, best_performance
def main(args):
# I/O
config_file = args.config_file
config = utils.import_file(config_file, 'config')
#trainset = utils.Dataset(config.train_dataset_path)
testset = utils.Dataset(config.test_dataset_path)
network = BaseNetwork()
network.initialize(config, 0 ) #trainset.num_classes
# Initalization for running
log_dir = utils.create_log_dir(config, config_file)
summary_writer = tf.summary.FileWriter(log_dir, network.graph)
if config.restore_model is not None:
network.restore_model(config.restore_model, config.restore_scopes)
# Set up LFW test protocol and load images
print('Loading images...')
lfwtest = LFWTest(testset.images)
lfwtest.init_standard_proto(config.lfw_pairs_file)
lfwtest.images = utils.preprocess(lfwtest.image_paths, config, is_training=False)
#trainset.start_batch_queue(config, True)
#
# Main Loop
#
print('\nStart Training\nname: %s\n# epochs: %d\nepoch_size: %d\nbatch_size: %d\n'\
% (config.name, config.num_epochs, config.epoch_size, config.batch_size))
global_step = 0
# Testing on LFW
print('Testing on Neetis LFW protocol...')
embeddings = network.extract_feature(lfwtest.images, config.batch_size)
print(type(embeddings))
accuracy_embeddings, threshold_embeddings = lfwtest.test_standard_proto(embeddings)
print('Embeddings Accuracy: %2.4f Threshold %2.3f' % (accuracy_embeddings, threshold_embeddings))
with open(os.path.join(log_dir,'lfw_result.txt'),'at') as f:
f.write('%d\t%.5f\n' % (global_step,accuracy_embeddings))
summary = tf.Summary()
summary.value.add(tag='lfw/accuracy', simple_value=accuracy_embeddings)
summary_writer.add_summary(summary, global_step)
def run():
args = argparser()
path = utils.create_log_dir(sys.argv)
utils.start(args.http_port)
env = Env(args)
agents = [Agent(args) for _ in range(args.n_agent)]
master = Master(args)
for agent in agents:
master.add_agent(agent)
master.add_env(env)
success_list = []
time_list = []
for idx in range(args.n_episode):
print('=' * 80)
print("Episode {}".format(idx + 1))
# 서버의 stack, timer 초기화
print("서버를 초기화하는중...")
master.reset(path)
# 에피소드 시작
master.start()
# 에이전트 학습
master.train()
print('=' * 80)
success_list.append(master.infos["is_success"])
time_list.append(master.infos["end_time"] - master.infos["start_time"])
if (idx + 1) % args.print_interval == 0:
print("=" * 80)
print("EPISODE {}: Avg. Success Rate / Time: {:.2} / {:.2}".format(
idx + 1, np.mean(success_list), np.mean(time_list)))
success_list.clear()
time_list.clear()
print("=" * 80)
if (idx + 1) % args.checkpoint_interval == 0:
utils.save_checkpoints(path, agents, idx + 1)
if args.visual:
visualize(path, args)
print("끝")
utils.close()
def main(args):
# I/O
config_file = args.config_file
config = utils.import_file(config_file, 'config')
trainset = utils.Dataset(config.train_dataset_path)
testset = utils.Dataset(config.test_dataset_path)
network = BaseNetwork()
network.initialize(config, trainset.num_classes)
# Initalization for running
log_dir = utils.create_log_dir(config, config_file)
summary_writer = tf.summary.FileWriter(log_dir, network.graph)
if config.restore_model is not None:
network.restore_model(config.restore_model, config.restore_scopes)
# Set up LFW test protocol and load images
print('Loading images...')
lfwtest = LFWTest(testset.images)
lfwtest.init_standard_proto(config.lfw_pairs_file)
lfwtest.images = utils.preprocess(lfwtest.image_paths,
config,
is_training=False)
trainset.start_batch_queue(config, True)
#
# Main Loop
#
print('\nStart Training\nname: %s\n# epochs: %d\nepoch_size: %d\nbatch_size: %d\n'\
% (config.name, config.num_epochs, config.epoch_size, config.batch_size))
global_step = 0
start_time = time.time()
for epoch in range(config.num_epochs):
# Training
for step in range(config.epoch_size):
# Prepare input
learning_rate = utils.get_updated_learning_rate(
global_step, config)
batch = trainset.pop_batch_queue()
wl, sm, global_step = network.train(batch['images'],
batch['labels'], learning_rate,
config.keep_prob)
# Display
if step % config.summary_interval == 0:
duration = time.time() - start_time
start_time = time.time()
utils.display_info(epoch, step, duration, wl)
summary_writer.add_summary(sm, global_step=global_step)
# Testing on LFW
print('Testing on standard LFW protocol...')
embeddings = network.extract_feature(lfwtest.images, config.batch_size)
accuracy_embeddings, threshold_embeddings = lfwtest.test_standard_proto(
embeddings)
print('Embeddings Accuracy: %2.4f Threshold %2.3f' %
(accuracy_embeddings, threshold_embeddings))
with open(os.path.join(log_dir, 'lfw_result.txt'), 'at') as f:
f.write('%d\t%.5f\n' % (global_step, accuracy_embeddings))
summary = tf.Summary()
summary.value.add(tag='lfw/accuracy', simple_value=accuracy_embeddings)
summary_writer.add_summary(summary, global_step)
# Save the model
network.save_model(log_dir, global_step)
help='Number of projections to test')
parser.add_argument('-e',
'--eps',
nargs='?',
default=0.001,
type=float,
help='Epsilon for entropic gw')
args = parser.parse_args()
all_samples = args.all_samples
numItermax = args.nitermax
projs = args.proj
epsilon = args.eps
niterentro = args.it_entro
log_dir = utils.create_log_dir(args)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
results = {}
results['all_samples'] = all_samples
results['projs'] = projs
results['epsilon'] = epsilon
results['all_gw'] = []
results['t_all_gw'] = []
results['all_converged'] = []
results['all_entro_gw'] = []
results['t_all_entro_gw'] = []
from os.path import join
import utils
import builders
import args
from core.base import Trainer
from core.plugins.logging import TrainLogger, MetricFileLogger, HeaderPrinter
from core.plugins.storage import BestModelSaver, ModelLoader
from metrics import KNNF1ScoreMetric, Evaluator
# Script arguments
args = args.get_args()
# Create directory to save plots, models, results, etc
log_path = utils.create_log_dir(args.exp_path, args)
print(f"Logging to {log_path}")
# Dump all script arguments
utils.dump_params(join(log_path, 'config.cfg'), args)
# Set custom seed
utils.set_custom_seed(args.seed)
# Load dataset and create model
print(f"[Model: {args.model.upper()}]")
print(f"[Loss: {args.loss.upper()}]")
print('[Loading Dataset and Model...]')
# Embedding dim is 768 based on BERT, we use the same for LSTM to be fair
# Classes are 6 because we include a 'non-misogyny class'
nfeat, nclass = 768, 6
config = builders.build_config(args, nfeat, nclass)
def main(args):
# I/O
config_file = args.config_file
config = utils.import_file(config_file, 'config')
trainset = utils.Dataset(config.train_dataset_path)
testset = utils.Dataset(config.test_dataset_path)
network = SiblingNetwork()
network.initialize(config, trainset.num_classes)
# Initalization for running
log_dir = utils.create_log_dir(config, config_file)
summary_writer = tf.summary.FileWriter(log_dir, network.graph)
if config.restore_model:
network.restore_model(config.restore_model, config.restore_scopes)
# Set up test protocol and load images
print('Loading images...')
testset.separate_template_and_probes()
testset.images = utils.preprocess(testset.images, config, is_training=False)
trainset.start_batch_queue(config, True)
#
# Main Loop
#
print('\nStart Training\nname: %s\n# epochs: %d\nepoch_size: %d\nbatch_size: %d\n'\
% (config.name, config.num_epochs, config.epoch_size, config.batch_size))
global_step = 0
start_time = time.time()
for epoch in range(config.num_epochs):
# Training
for step in range(config.epoch_size):
# Prepare input
learning_rate = utils.get_updated_learning_rate(global_step, config)
image_batch, label_batch = trainset.pop_batch_queue()
switch_batch = utils.zero_one_switch(len(image_batch))
wl, sm, global_step = network.train(image_batch, label_batch, switch_batch, learning_rate, config.keep_prob)
# Display
if step % config.summary_interval == 0:
duration = time.time() - start_time
start_time = time.time()
utils.display_info(epoch, step, duration, wl)
summary_writer.add_summary(sm, global_step=global_step)
# Testing
print('Testing...')
switch = utils.zero_one_switch(len(testset.images))
embeddings = network.extract_feature(testset.images, switch, config.batch_size)
tars, fars, _ = utils.test_roc(embeddings, FARs=[1e-4, 1e-3, 1e-2])
with open(os.path.join(log_dir,'result.txt'),'at') as f:
for i in range(len(tars)):
print('[%d] TAR: %2.4f FAR %2.3f' % (epoch+1, tars[i], fars[i]))
f.write('[%d] TAR: %2.4f FAR %2.3f\n' % (epoch+1, tars[i], fars[i]))
summary = tf.Summary()
summary.value.add(tag='test/tar_%d'%i, simple_value=tars[i])
summary_writer.add_summary(summary, global_step)
# Save the model
network.save_model(log_dir, global_step)
parser.add_argument('--config_file',
help='Path to training configuration file',
type=str)
config_file = parser.parse_args().config_file
# I/O
config = utils.import_file(config_file, 'config')
trainset = utils.Dataset(config.splits_path + '/train_' +
str(config.fold_number) + '.txt')
trainset.images = utils.preprocess(trainset.images, config, True)
network = Network()
network.initialize(config, trainset.num_classes)
# Initalization for running
log_dir = utils.create_log_dir(config, config_file)
summary_writer = tf.summary.FileWriter(log_dir, network.graph)
if config.restore_model:
network.restore_model(config.restore_model, config.restore_scopes)
# Load gallery and probe file_list
print('Loading images...')
probes = []
gal = []
with open(
config.splits_path + '/fold_' + str(config.fold_number) +
'/probe_1.txt', 'r') as f:
for line in f:
probes.append(line.strip())
probe_set = evaluate.ImageSet(probes, config)
if args.alpha==-8000 and not args.automatic_cv_alpha:
raise AlphaMustBeDefinedError('You must set alpha via -a or use automatic grid via -cva')
name='fgw'+'_'+args.dataset_name+'_feature_metric_'+args.feature_metric+'_structure_metric_'+args.structure_metric
if args.wl_feature>0:
name=name+'_wl_'+str(args.wl_feature)
name=name+args.optionnal_name
try:
if not os.path.exists(args.log_dir):
os.makedirs(args.log_dir)
except OSError:
raise
log_dir=create_log_dir(args)
if args.Csvm !=-1:
Clist=[args.Csvm]
else:
Clist=list(np.logspace(-4,4,15))
if args.automatic_cv_alpha :
nb=15
N=int(15/3)
a=np.logspace(-6,-1,N)
c=1-a
b=np.array(list(set(np.linspace(a[0],c[0],N)).difference(set((a[0],c[0])))))
alphalist=np.concatenate((a,b,c))
alpha_list=list(np.sort(np.append([0,1],alphalist)))
else:
def main(_):
# We want to see all the logging messages for this tutorial.
# 记录日志
tf.logging.set_verbosity(tf.logging.INFO)
# Start a new TensorFlow session.
# 启动会话
sess = tf.InteractiveSession()
# 将日志写入指定文件
# get TF logger
import utils
utils.create_log_dir(FLAGS.train_dir, FLAGS.log_path)
# Begin by making sure we have the training data we need. If you already have
# training data of your own, use `--data_url= ` on the command line to avoid
# downloading.
# 模型设置：单词切割、采样率、时长、窗长、帧移、mfcc系数
model_settings = models.prepare_model_settings(
len(input_data.prepare_words_list(FLAGS.wanted_words.split(','))),
FLAGS.sample_rate, FLAGS.clip_duration_ms, FLAGS.window_size_ms,
FLAGS.window_stride_ms, FLAGS.dct_coefficient_count)
# 数据网址、本地文件夹、静音比重、干扰词比重、验证集占比、测试集占比
audio_processor = input_data.AudioProcessor(FLAGS.data_url, FLAGS.data_dir,
FLAGS.silence_percentage,
FLAGS.unknown_percentage,
FLAGS.wanted_words.split(','),
FLAGS.validation_percentage,
FLAGS.testing_percentage,
model_settings)
# 数据输入格式
fingerprint_size = model_settings['fingerprint_size']
# 标签数量
label_count = model_settings['label_count']
# 采样率
time_shift_samples = int((FLAGS.time_shift_ms * FLAGS.sample_rate) / 1000)
# Figure out the learning rates for each training phase. Since it's often
# effective to have high learning rates at the start of training, followed by
# lower levels towards the end, the number of steps and learning rates can be
# specified as comma-separated lists to define the rate at each stage. For
# example --how_many_training_steps=10000,3000 --learning_rate=0.001,0.0001
# will run 13,000 training loops in total, with a rate of 0.001 for the first
# 10,000, and 0.0001 for the final 3,000.
# 实验发现：在训练的开始，能很快到达很高的准确率，即训练速度快；但在接近最高点的时候，学习效率就会降低；
# 因此为了适应前后训练的不同，需要根据时间调整前后的学习率：前面大，后面小；
# 比如整个循环有13000次，前10000次使用0.001的学习率；后3000次使用0.0001的学习率
# training_steps_list=10000,10000,10000
# learning_rates_list=0.0005, 0.0001, 0.00002
training_steps_list = list(
map(int, FLAGS.how_many_training_steps.split(',')))
learning_rates_list = list(map(float, FLAGS.learning_rate.split(',')))
if len(training_steps_list) != len(learning_rates_list):
raise Exception(
'--how_many_training_steps and --learning_rate must be equal length '
'lists, but are %d and %d long instead' %
(len(training_steps_list), len(learning_rates_list)))
# 数据输入占位符
fingerprint_input = tf.placeholder(tf.float32, [None, fingerprint_size],
name='fingerprint_input')
# 创建模型
# logits:预测标签结果;dropout_prob:丢弃率
logits, dropout_prob = models.create_model(fingerprint_input,
model_settings,
FLAGS.model_architecture,
FLAGS.model_size_info,
is_training=True)
# Define loss and optimizer
# 输出维度占位符---真实标签
ground_truth_input = tf.placeholder(tf.float32, [None, label_count],
name='groundtruth_input')
# Optionally we can add runtime checks to spot when NaNs or other symptoms of
# numerical errors start occurring during training.
control_dependencies = []
if FLAGS.check_nans:
checks = tf.add_check_numerics_ops()
control_dependencies = [checks]
# Create the back propagation and training evaluation machinery in the graph.
# 交叉验证，即优化函数
with tf.name_scope('cross_entropy'):
cross_entropy_mean = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(labels=ground_truth_input,
logits=logits))
# summary概要：主要用于数据可视化，scalar将数据转化为标量
tf.summary.scalar('cross_entropy', cross_entropy_mean)
# 使用BN，即批量更新
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
with tf.name_scope('train'), tf.control_dependencies(
update_ops), tf.control_dependencies(control_dependencies):
# 学习率
learning_rate_input = tf.placeholder(tf.float32, [],
name='learning_rate_input')
# 优化函数
train_op = tf.train.AdamOptimizer(learning_rate_input)
# 交叉验证函数+优化函数 进行训练
train_step = slim.learning.create_train_op(cross_entropy_mean,
train_op)
# train_step = tf.train.GradientDescentOptimizer(
# learning_rate_input).minimize(cross_entropy_mean)
# 返回概率最大的标签下标：分别为预测值-真实值
predicted_indices = tf.argmax(logits, 1)
expected_indices = tf.argmax(ground_truth_input, 1)
# 正确与否，0-1值矩阵
correct_prediction = tf.equal(predicted_indices, expected_indices)
# 计算混淆矩阵：直接显示模型的效果，越集中于对角线，表示效果越好
# https: // www.zhihu.com / question / 36883196
confusion_matrix = tf.confusion_matrix(expected_indices,
predicted_indices,
num_classes=label_count)
# 正确率：计算平均值，因为由0-1组成，所以平均值即正确率
evaluation_step = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
tf.summary.scalar('accuracy', evaluation_step)
# 全局步数记录
global_step = tf.train.get_or_create_global_step()
# 全局步数+1
increment_global_step = tf.assign(global_step, global_step + 1)
# 模型存储
saver = tf.train.Saver(tf.global_variables())
# Merge all the summaries and write them out to /Users/zoutai/ML_KWS/retrain_logs (by default)
# 将所有的数据概要保存到文件
merged_summaries = tf.summary.merge_all()
train_writer = tf.summary.FileWriter(FLAGS.summaries_dir + '/train',
sess.graph)
validation_writer = tf.summary.FileWriter(FLAGS.summaries_dir +
'/validation')
# 初始化全局变量
tf.global_variables_initializer().run()
# Parameter counts
# 记录参数量
params = tf.trainable_variables()
num_params = sum(map(lambda t: np.prod(tf.shape(t.value()).eval()),
params))
# print('Total number of Parameters: ', num_params)
tf.logging.info('Total number of Parameters: %d ', num_params)
start_step = 1
# 是否有初始模型，如果有从初始模型开始；否则，重新开始训练
if FLAGS.start_checkpoint:
models.load_variables_from_checkpoint(sess, FLAGS.start_checkpoint)
start_step = global_step.eval(session=sess)
start_step = start_step // 2
tf.logging.info('Training from step: %d ', start_step)
# Save graph.pbtxt.
# 存储模型
tf.train.write_graph(sess.graph_def, FLAGS.train_dir,
FLAGS.model_architecture + '.pbtxt')
# Save list of words.
# 记录标签
with gfile.GFile(
os.path.join(FLAGS.train_dir,
FLAGS.model_architecture + '_labels.txt'), 'w') as f:
f.write('\n'.join(audio_processor.words_list))
# 开启训练
# Training loop.
best_accuracy = 0
# 总训练步数
training_steps_max = np.sum(training_steps_list)
for training_step in range(start_step, training_steps_max + 1):
# Figure out what the current learning rate is.
# 每一阶段的总步数
training_steps_sum = 0
# 三个阶段
for i in range(len(training_steps_list)):
training_steps_sum += training_steps_list[i]
if training_step <= training_steps_sum:
learning_rate_value = learning_rates_list[i]
break
# Pull the audio samples we'll use for training.
# 读取输入数据
train_fingerprints, train_ground_truth = audio_processor.get_data(
FLAGS.batch_size, 0, model_settings, FLAGS.background_frequency,
FLAGS.background_volume, time_shift_samples, 'training', sess)
# 1.训练部分
# Run the graph with this batch of training data.
# 开启训练会话
train_summary, train_accuracy, cross_entropy_value, _, _ = sess.run(
[
merged_summaries, evaluation_step, cross_entropy_mean,
train_step, increment_global_step
],
feed_dict={
fingerprint_input: train_fingerprints,
ground_truth_input: train_ground_truth,
learning_rate_input: learning_rate_value,
dropout_prob: 1.0
})
train_writer.add_summary(train_summary, training_step)
tf.logging.info(
'Step #%d: rate %f, accuracy %.2f%%, cross entropy %f' %
(training_step, learning_rate_value, train_accuracy * 100,
cross_entropy_value))
# 2.验证部分
# 最后一步，训练完成
is_last_step = (training_step == training_steps_max)
# 达到一定步数进行一次记录，一般为500次；或者到达最终步数记录一次
if (training_step % FLAGS.eval_step_interval) == 0 or is_last_step:
# 读取验证集，进行验证
set_size = audio_processor.set_size('validation')
total_accuracy = 0
total_conf_matrix = None
for i in range(0, set_size, FLAGS.batch_size):
validation_fingerprints, validation_ground_truth = (
audio_processor.get_data(FLAGS.batch_size, i,
model_settings, 0.0, 0.0, 0,
'validation', sess))
# Run a validation step and capture training summaries for TensorBoard
# with the `merged` op.
validation_summary, validation_accuracy, conf_matrix = sess.run(
[merged_summaries, evaluation_step, confusion_matrix],
feed_dict={
fingerprint_input: validation_fingerprints,
ground_truth_input: validation_ground_truth,
dropout_prob: 1.0
})
validation_writer.add_summary(validation_summary,
training_step)
batch_size = min(FLAGS.batch_size, set_size - i)
total_accuracy += (validation_accuracy * batch_size) / set_size
if total_conf_matrix is None:
total_conf_matrix = conf_matrix
else:
total_conf_matrix += conf_matrix
tf.logging.info('Confusion Matrix:\n %s' % (total_conf_matrix))
tf.logging.info('Step %d: Validation accuracy = %.2f%% (N=%d)' %
(training_step, total_accuracy * 100, set_size))
# Save the model checkpoint when validation accuracy improves
# 存储当前模型
if total_accuracy > best_accuracy:
best_accuracy = total_accuracy
checkpoint_path = os.path.join(
FLAGS.train_dir, 'best', FLAGS.model_architecture + '_' +
str(int(best_accuracy * 10000)) + '.ckpt')
tf.logging.info('Saving best model to "%s-%d"',
checkpoint_path, training_step)
saver.save(sess, checkpoint_path, global_step=training_step)
tf.logging.info('So far the best validation accuracy is %.2f%%' %
(best_accuracy * 100))
# 3.测试部分
set_size = audio_processor.set_size('testing')
tf.logging.info('set_size=%d', set_size)
total_accuracy = 0
total_conf_matrix = None
for i in range(0, set_size, FLAGS.batch_size):
test_fingerprints, test_ground_truth = audio_processor.get_data(
FLAGS.batch_size, i, model_settings, 0.0, 0.0, 0, 'testing', sess)
test_accuracy, conf_matrix = sess.run(
[evaluation_step, confusion_matrix],
feed_dict={
fingerprint_input: test_fingerprints,
ground_truth_input: test_ground_truth,
dropout_prob: 1.0
})
batch_size = min(FLAGS.batch_size, set_size - i)
total_accuracy += (test_accuracy * batch_size) / set_size
if total_conf_matrix is None:
total_conf_matrix = conf_matrix
else:
total_conf_matrix += conf_matrix
tf.logging.info('Confusion Matrix:\n %s' % (total_conf_matrix))
tf.logging.info('Final test accuracy = %.2f%% (N=%d)' %
(total_accuracy * 100, set_size))
def train(config_file, counter):
# I/O
config = utils.import_file(config_file, 'config')
splits_path = config.splits_path + '/split{}'.format(counter)
trainset = utils.Dataset(splits_path + '/train_' + str(config.fold_number) + '.txt')
trainset.images = utils.preprocess(trainset.images, config, True)
network = Network()
network.initialize(config, trainset.num_classes)
# Initalization for running
log_dir = utils.create_log_dir(config, config_file)
summary_writer = tf.compat.v1.summary.FileWriter(log_dir, network.graph)
if config.restore_model:
network.restore_model(config.restore_model, config.restore_scopes)
# Load gallery and probe file_list
print('Loading images...')
probes = []
gal = []
with open(splits_path + '/fold_' + str(config.fold_number) + '/probe_1.txt' ,'r') as f:
for line in f:
probes.append(line.strip())
probe_set = evaluate.ImageSet(probes, config)
#probe_set.extract_features(network, len(probes))
#
with open(splits_path + '/fold_'+ str(config.fold_number) + '/gal_1.txt', 'r') as f:
for line in f:
gal.append(line.strip())
gal_set = evaluate.ImageSet(gal, config)
#gal_set.extract_features(network, len(gal))
trainset.start_batch_queue(config, True)
#
# Main Loop
#
print('\nStart Training\n# epochs: {}\nepoch_size: {}\nbatch_size: {}\n'.\
format(config.num_epochs, config.epoch_size, config.batch_size))
global_step = 0
start_time = time.time()
for epoch in range(config.num_epochs):
# Training
for step in range(config.epoch_size):
# Prepare input
learning_rate = utils.get_updated_learning_rate(global_step, config)
image_batch, label_batch = trainset.pop_batch_queue()
wl, sm, global_step = network.train(image_batch, label_batch, learning_rate, config.keep_prob)
# Display
if step % config.summary_interval == 0:
# visualize.scatter2D(_prelogits[:,:2], _label_batch, _pgrads[0][:,:2])
duration = time.time() - start_time
start_time = time.time()
utils.display_info(epoch, step, duration, wl)
summary_writer.add_summary(sm, global_step=global_step)
# Testing
print('Testing...')
probe_set.extract_features(network, len(probes))
gal_set.extract_features(network, len(gal))
rank1, rank5 = evaluate.identify(log_dir, probe_set, gal_set)
print('rank-1: {:.3f}, rank-5: {:.3f}'.format(rank1[0], rank5[0]))
# Output test result
summary = tf.Summary()
summary.value.add(tag='identification/rank1', simple_value=rank1[0])
summary.value.add(tag='identification/rank5', simple_value=rank5[0])
summary_writer.add_summary(summary, global_step)
# Save the model
network.save_model(log_dir, global_step)
results_copy = os.path.join('log/result_{}_{}.txt'.format(config.model_version, counter))
shutil.copyfile(os.path.join(log_dir,'result.txt'), results_copy)
