def view_tongue_data_augment():
patient_tongue_dir = config['root_path'] + \
config['original_path'] + 'tongue_9585'
tongue_zhiliao_path = config['root_path'] + \
config['original_path'] + 'tongue_zhiliao.list'
yaopin_path = config['root_path'] + \
config['original_path'] + 'yaopin.vocab'
tongue_ids, tongue_image_arrays, tongue_yaofangs, tongue_image_shape = patient_tongue_generator.loadDatafromFile(
patient_tongue_dir, tongue_zhiliao_path, image_normal_size=(224, 224))
# fetch max(id) in yaopin.vocab as nb_yao
with open(yaopin_path, 'r') as yaopin_file:
nb_yao = max(
int(line.split(' ')[0]) for line in yaopin_file.readlines())
total_tongue_x, total_y = tongue2text_gen.data_tensorization(
tongue_image_arrays, tongue_yaofangs, tongue_image_shape, nb_yao)
datagen = image_augment.image_augment_gen()
augmented_x, augmented_y = image_augment.data_tensoration_augment(
datagen, total_tongue_x, total_y)
print(np.shape(augmented_x))
print(np.shape(augmented_y))
def gen_deeper_pipeline_predictor_test(tongue_image_arrays,
tongue_yaofangs,
tongue_image_shape,
nb_yao,
trained_gen_model,
use_tfidf_tensor=False):
'''
@param use_tfidf_tensor: flag of use tfidf tensor or not with different tensorization function
'''
''' load test_x & test_y '''
if use_tfidf_tensor == True:
total_x, total_y = tongue2text_gen.data_tensorization_tfidf(
tongue_image_arrays, tongue_yaofangs, tongue_image_shape, nb_yao)
else:
total_x, total_y = tongue2text_gen.data_tensorization(
tongue_image_arrays, tongue_yaofangs, tongue_image_shape, nb_yao)
# test_x = total_x[: 500]
# test_x = total_x[2000 : 2500]
# test_x = total_x[4000 : 4500]
# test_x = total_x[6000 : 6500]
test_x = total_x[len(total_x) - 500:]
gen_output = tongue2text_gen.predictor(trained_gen_model, test_x)
return gen_output
def tongue_basic_gen_trainer(tongue_image_arrays, tongue_yaofangs, tongue_image_shape, nb_yao,
gen_model_path=None, train_on_batch=False,
use_data_augment=False):
'''
@param use_tfidf_tensor: flag of use tfidf tensor or not with different tensorization function
'''
total_tongue_x, total_y = tongue2text_gen.data_tensorization(
tongue_image_arrays, tongue_yaofangs, tongue_image_shape, nb_yao)
# train_x = total_tongue_x[500:]
# train_y = total_y[500:]
# train_x = np.concatenate((total_tongue_x[: 2000], total_tongue_x[2500 :]), axis=0)
# train_y = np.concatenate((total_y[: 2000], total_y[2500 :]), axis=0)
# train_x = np.concatenate((total_tongue_x[: 4000], total_tongue_x[4500 :]), axis=0)
# train_y = np.concatenate((total_y[: 4000], total_y[4500 :]), axis=0)
# train_x = np.concatenate((total_tongue_x[: 6000], total_tongue_x[6500 :]), axis=0)
# train_y = np.concatenate((total_y[: 6000], total_y[6500 :]), axis=0)
train_x = total_tongue_x[: len(total_tongue_x) - 500]
train_y = total_y[: len(total_y) - 500]
del(total_tongue_x)
del(total_y)
if use_data_augment == True:
# just can be use on service 225 with big memory
datagen = image_augment.image_augment_gen()
train_x, train_y = image_augment.data_tensoration_augment(
datagen, train_x, train_y)
scaling_act_type = 'binary'
print('training 2 * cnn + mlp tongue2text gen model------on_batch: %d------scaling_activation: %s...' %
(train_on_batch, scaling_act_type))
tongue_gen_model = tongue2text_gen.k_cnns_mlp(
yao_indices_dim=nb_yao, tongue_image_shape=tongue_image_shape, with_compile=True)
if train_on_batch == True:
trained_tongue_gen_model, history = tongue2text_gen.trainer_on_batch(
tongue_gen_model, train_x, train_y)
else:
record_path = None
if gen_model_path != None:
record_path = gen_model_path.replace('json', 'h5')
trained_tongue_gen_model, history = tongue2text_gen.trainer(
tongue_gen_model, train_x, train_y, best_record_path=record_path)
if gen_model_path != None:
tongue2text_gen.storageModel(model=trained_tongue_gen_model, frame_path=gen_model_path,
replace_record=False)
print('history: {0}'.format(history))
return trained_tongue_gen_model
def tongue_gen_deeper_1pipeline_trainer(tongue_image_arrays,
tongue_yaofangs,
tongue_image_shape,
base_model_name,
nb_yao,
gen_model_path=None,
train_on_batch=False,
use_tfidf_tensor=False,
use_data_augment=True):
'''
@param base_model: the name of base model in {'vgg16', 'vgg19', 'resnet50'}
'''
total_tongue_x, total_y = tongue2text_gen.data_tensorization(
tongue_image_arrays, tongue_yaofangs, tongue_image_shape, nb_yao)
# train_x = total_tongue_x[500:]
# train_y = total_y[500:]
# train_x = np.concatenate((total_tongue_x[: 2000], total_tongue_x[2500 :]), axis=0)
# train_y = np.concatenate((total_y[: 2000], total_y[2500 :]), axis=0)
# train_x = np.concatenate((total_tongue_x[: 4000], total_tongue_x[4500 :]), axis=0)
# train_y = np.concatenate((total_y[: 4000], total_y[4500 :]), axis=0)
# train_x = np.concatenate((total_tongue_x[: 6000], total_tongue_x[6500 :]), axis=0)
# train_y = np.concatenate((total_y[: 6000], total_y[6500 :]), axis=0)
train_x = total_tongue_x[:len(total_tongue_x) - 500]
train_y = total_y[:len(total_y) - 500]
del (total_tongue_x)
del (total_y)
if use_data_augment == True:
# just can be use on service 225 with big memory
datagen = image_augment.image_augment_gen()
train_x, train_y = image_augment.data_tensoration_augment(
datagen, train_x, train_y)
scaling_act_type = 'tfidf' if use_tfidf_tensor else 'binary'
print(
'training 1pipeline + mlp tongue2text gen model------on_batch: %d------scaling_activation: %s...'
% (train_on_batch, scaling_act_type))
print('use base_model: ' + base_model_name)
image_input, base_model = tongue2text_deeper_gen.k_base_model(
tongue_image_shape=tongue_image_shape, model_name=base_model_name)
tongue_gen_model = tongue2text_deeper_gen.k_1pipeline_mlp(
yao_indices_dim=nb_yao,
image_input=image_input,
base_model=base_model,
with_compile=True)
if train_on_batch == True:
trained_tongue_gen_model, history = tongue2text_gen.trainer_on_batch(
tongue_gen_model, train_x, train_y)
else:
record_path = None
if gen_model_path != None:
record_path = gen_model_path.replace('json', 'h5')
trained_tongue_gen_model, history = tongue2text_gen.trainer(
tongue_gen_model,
train_x,
train_y,
batch_size=16,
epochs=_new_training_epochs,
best_record_path=record_path)
if gen_model_path != None:
tongue2text_gen.storageModel(model=trained_tongue_gen_model,
frame_path=gen_model_path,
replace_record=False)
print('history: {0}'.format(history))
return trained_tongue_gen_model