def prepare_multi_modal_data(files_path, task: Task, images_size=(128, 128), with_split=True):
path = os.path.join(str(fedot_project_root()), files_path)
unpack_archived_data(path)
data = InputData.from_json_files(path, fields_to_use=['votes', 'year'],
label='rating', task=task)
class_labels = np.asarray([0 if t <= 7 else 1 for t in data.target])
data.target = class_labels
ratio = 0.5
img_files_path = f'{files_path}/*.jpeg'
img_path = os.path.join(str(fedot_project_root()), img_files_path)
data_img = InputData.from_image(images=img_path, labels=class_labels, task=task, target_size=images_size)
data_text = InputData.from_json_files(path, fields_to_use=['plot'],
label='rating', task=task,
data_type=DataTypesEnum.text)
data_text.target = class_labels
if with_split:
train_num, test_num = train_test_data_setup(data, shuffle_flag=False, split_ratio=ratio)
train_img, test_img = train_test_data_setup(data_img, shuffle_flag=False, split_ratio=ratio)
train_text, test_text = train_test_data_setup(data_text, shuffle_flag=False, split_ratio=ratio)
else:
train_num, test_num = data, data
train_img, test_img = data_img, data_img
train_text, test_text = data_text, data_text
return train_num, test_num, train_img, test_img, train_text, test_text
def run_image_classification_problem(train_dataset: tuple,
test_dataset: tuple,
composite_flag: bool = True):
task = Task(TaskTypesEnum.classification)
x_train, y_train = train_dataset[0], train_dataset[1]
x_test, y_test = test_dataset[0], test_dataset[1]
dataset_to_train = InputData.from_image(images=x_train,
labels=y_train,
task=task)
dataset_to_validate = InputData.from_image(images=x_test,
labels=y_test,
task=task)
chain = get_composite_chain(composite_flag)
chain.fit(input_data=dataset_to_train)
predictions = chain.predict(dataset_to_validate)
roc_auc_on_valid = calculate_validation_metric(predictions,
dataset_to_validate)
print(f'ROCAUC: {roc_auc_on_valid}')
return roc_auc_on_valid, dataset_to_train, dataset_to_validate
