def pre_val_to_train(self, train_vect_in_path, val_in_path):
"""
:param train_vect_in_path:
:param val_in_path:
:return:
"""
train_vect = utils.load_pickle(train_vect_in_path)
val_vect = utils.load_pickle(val_in_path)
n = 1280
train_vect = train_vect + val_vect[n:]
val_vect = val_vect[:n]
utils.save_pickle(train_vect, train_vect_in_path)
utils.save_pickle(val_vect, val_in_path)
def fetch_val_batch(self, caption_data_path, image_data_path, batch_size):
"""
:param caption_data_path:
:param image_data_path:
:param batch_size:
:return:
"""
captions = utils.load_pickle(caption_data_path)
val_data_count = len(captions)
if TrainingArg.check_test:
val_data_count = batch_size
print('WARNING: CHECK TEST MODEL NOW')
start = 0
end = batch_size
for item in range(0, val_data_count, batch_size):
temp = captions[start:end]
caption_batch = []
image_batch = []
for x in temp:
caption_batch.append(x['caption'])
image_batch.append(x['image_id'])
try:
image_batch = self.pre_data_with_vgg19(image_data_path,
image_batch)
except Exception as e:
print('load image failed: {}-{}, cause {}'.format(
start, end, e))
start += batch_size
end += batch_size
continue
start += batch_size
end += batch_size
yield caption_batch, image_batch
def __init__(self):
self.word_to_idx = utils.load_pickle(Const.vocab_path)
self.model = CaptionGenerator(self.word_to_idx,
dim_feature=[196, 512],
dim_embed=512,
dim_hidden=1024,
n_time_step=33,
prev2out=True,
ctx2out=True,
alpha_c=1.0,
selector=True,
dropout=True)
self.n_epochs = TrainingArg.n_epochs
self.batch_size = TrainingArg.batch_size
self.update_rule = TrainingArg.update_rule
self.learning_rate = TrainingArg.learning_rate
self.print_bleu = TrainingArg.print_bleu
self.print_every = TrainingArg.print_every
self.save_every = TrainingArg.save_every
self.log_path = TrainingArg.log_path
self.model_path = TrainingArg.model_path
self.pretrained_model = TrainingArg.pretrained_model
self.test_model = TrainingArg.test_model
self.max_words_len = 35
self.pre_mgr = PreData(vgg19_path=TrainingArg.vgg19_path) # 数据管理
def __init__(self):
self.word_to_idx = utils.load_pickle(Const.vocab_path)
self.model = CaptionGenerator(self.word_to_idx,
dim_feature=[196, 512],
dim_embed=512,
dim_hidden=1024,
n_time_step=33,
prev2out=True,
ctx2out=True,
alpha_c=1.0,
selector=True,
dropout=True)
self.n_epochs = TrainingArg.n_epochs
self.batch_size = TrainingArg.batch_size
self.update_rule = TrainingArg.update_rule
self.learning_rate = TrainingArg.learning_rate
self.print_bleu = TrainingArg.print_bleu
self.print_every = TrainingArg.print_every
self.save_every = TrainingArg.save_every
self.log_path = TrainingArg.log_path # FLAGS.log_dir
self.model_path = TrainingArg.model_path # FLAGS.output_dir # TrainingArg.model_path
self.data_dir = Const.resize_train_out_path # FLAGS.data_dir
self.pretrained_model = TrainingArg.pretrained_model
self.test_model = TrainingArg.test_model
self.max_words_len = 35
self.pre_mgr = PreData(vgg19_path=TrainingArg.vgg19_path) # 数据管理
# set an optimizer by update rule
if self.update_rule == 'adam':
self.optimizer = tf.train.AdamOptimizer
elif self.update_rule == 'momentum':
self.optimizer = tf.train.MomentumOptimizer
elif self.update_rule == 'rmsprop':
self.optimizer = tf.train.RMSPropOptimizer
if not os.path.exists(self.model_path):
os.makedirs(self.model_path)
if not os.path.exists(self.log_path):
os.makedirs(self.log_path)
self.org_decoded = {}
self.val_data_flag = False
def fetch_batch(self, caption_data_path, image_data_path, batch_size,
epochs):
"""
:param caption_data_path:
:param image_data_path:
:param batch_size:
:param epochs:
:return:
"""
captions = utils.load_pickle(caption_data_path)
self.train_data_count = len(captions)
for epoch in range(4, epochs):
start = 0
end = batch_size
random.shuffle(captions)
for item in range(0, self.train_data_count, batch_size):
temp = captions[start:end]
random.shuffle(temp)
if epoch < 5:
cap_index = epoch % 5
else:
cap_index = random.randint(0, 4)
caption_batch = []
image_batch = []
for x in temp:
caption_batch.append(x['caption'][cap_index])
image_batch.append(x['image_id'])
try:
image_batch = self.pre_data_with_vgg19(
image_data_path, image_batch)
except Exception as e:
print('load image failed: {}-{}, cause {}'.format(
start, end, e))
start += batch_size
end += batch_size
continue
start += batch_size
end += batch_size
yield caption_batch, image_batch, epoch
def load_vocab(self, vocab_path):
"""
:return:
"""
return utils.load_pickle(vocab_path)