def cli():
if args.recommend:
recommendation.start(args.meme)
if args.force_index:
index_data.start(args.force_index)
if args.search:
searchp.start(args)
if args.preprocess:
preprocess.start(args)
if args.generate:
meme_generator.start(args)
def cli():
#uses command line args to invoke sevices
if args.recommend:
recommendation.start(args.meme)
if args.force_index:
index_data.start(args.force_index)
if args.search:
searchp.start(args)
if args.preprocess:
preprocess.start(args)
if args.generate:
meme_generator.start(args)
def run_experiment(args):
"""run training and save to wandb"""
wandb.init(project=args.project_name)
sample_rate = args.sampling_rate
X, y_labels = preprocess.start("Intra",
sample_rate,
test_path_folder=None,
generator=False)
model = build_model(X[0].shape,
conv_1_size=args.conv_1_size,
batch_norm=args.batch_norm,
conv_2_size=args.conv_2_size,
kernel_size=args.kernel_size,
layer_1_size=args.layer_1_size,
dropout=args.dropout,
layer_2_size=args.layer_2_size,
optimizer=args.optimizer,
learning_rate=args.learning_rate,
weight_initializers=args.weight_initializers)
# log all values of interest to wandb
wandb.config.update(args)
dictionary_metrics = {out: [] for i, out in enumerate(model.metrics_names)}
skf = StratifiedKFold(n_splits=5)
for train_index, test_index in skf.split(X, y_labels):
y = preprocess.convert_y_labels_to_hot_vectors(y_labels)
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
model.fit(X_train,
y_train,
epochs=args.epochs,
batch_size=args.batch_size,
verbose=1)
metrics = model.evaluate(X_test, y_test)
for i, metrics_names in enumerate(model.metrics_names):
dictionary_metrics[metrics_names].append(metrics[i])
for key in dictionary_metrics:
dictionary_metrics[key] = np.array(dictionary_metrics[key]).mean()
wandb.log(dictionary_metrics)
def run_experiment(args):
"""run training and save to wandb"""
wandb.init(project=args.project_name)
sample_rate = args.sampling_rate
X, y_labels = preprocess.start(
"Cross", sample_rate, test_path_folder=None, generator=False)
X_train, X_val, y_train, y_val = train_test_split(
X, y_labels, test_size=0.06, random_state=42, stratify=y_labels)
y_train = preprocess.convert_y_labels_to_hot_vectors(y_train)
y_val = preprocess.convert_y_labels_to_hot_vectors(y_val)
X_train = preprocess.outliers_removal(X_train)
# First head configuration
kernel_size_head_1 = int(X_train[1] * 0.05)
strides_head_1 = int(kernel_size_head_1/2)
# Second head configuration
kernel_size_head_2 = int(X_train[1] * 0.25)
strides_head_2 = int(kernel_size_head_2/2)
wandb.config.update(args)
indices_train = np.arange(X_train.shape[0])
indices_val = np.arange(X_val.shape[0])
np.random.shuffle(indices_train)
np.random.shuffle(indices_val)
X_train = X_train[indices_train]
y_train = y_train[indices_train]
X_val = X_train[indices_val]
y_val = y_train[indices_val]
model = build_model(X_train, y_train, X_val, y_val, conv_1_size=args.conv_1_size, conv_2_size=args.conv_2_size, kernel_size_head_1=kernel_size_head_1, strides_head_1=strides_head_1, kernel_size_head_2=kernel_size_head_2,
strides_head_2=strides_head_2, layer_1_size=args.layer_1_size, dropout_conv=args.dropout_conv, dropout_dense=args.dropout_dense, layer_2_size=args.layer_2_size, optimizer=args.optimizer, learning_rate=args.learning_rate, weight_initializers=args.weight_initializers)
# log all values of interest to wandb
model.fit([X_train, X_train], y_train, validation_data=([X_val, X_val], y_val), epochs=args.epochs,
batch_size=args.batch_size, callbacks=[WandbCallback(monitor="val_loss", mode="min")], verbose=1)
camera.capture(image_path[count])
print("mengambil gambar", count+1, " selesai")
GPIO.output(flash, GPIO.LOW)
count += 1
start = False
print('Mengambil gambar selesai')
bunyi(buzz, 2)
stop = False
waktu_mulai = time.clock()
pic = 0
text = ""
while pic < count:
im = Image.open(image_path[pic])
im = preprocess.start(im)
temp = pytesseract.image_to_string(im, lang='ind', config='--psm 6')
temp = temp.replace('-', ' ')
temp = temp.replace('\n', ' ')
temp = re.sub(r'[^a-zA-Z0-9 ]','', temp)
text = text + temp
#im.close()
pic += 1
print(text)
bunyi(buzz, 3)
print("INFO: Saying Text")
print('Total huruf : ', len(text))
selesai = time.clock() - waktu_mulai
feature_kappas_LDAC[feature] = calculate_kappa(grade_dict,
LDAC_grades[feature])
return (feature_kappas_SVM, feature_kappas_LDAC)
if __name__ == '__main__':
global kappa_all_features
#if not path.exists("cache\\SVM_features_kappas.pickle") or not path.exists("cache\\LDAC_features_kappas.pickle"):
# experiment.start_experiment(teacher_grades=grade_dict, essays=essay_collection)
if not path.exists("cache\\SVM_model.pickle") or not path.exists(
"cache\\LDAC_model.pickle"):
if path.exists("cache\\essays.pickle"):
essay_collection = cache_manager.read_essays()
else:
essay_collection = preprocess.start()
cache_manager.cache_essays(essay_collection)
if path.exists("cache\\grades.pickle"):
grade_dict = cache_manager.read_grades()
else:
grade_dict = cache_manager.input_grades(essay_collection)
train_all_features()
feature_grades = train_each_feature()
feature_kappas = evaluate_each_feature(feature_grades[0],
feature_grades[1])
printing.print_feature_kappas(feature_kappas[0], feature_kappas[1])
else:
svm = cache_manager.read_classifier()
ldac = cache_manager.read_classifier("LDAC_model.pickle")
essay_vector = preprocess.start(normalise=True, path="essays_to_test/")
if (len(essay_vector) >= 1):