"""
run this file to generate the yolo format dataset
"""
import os
import yaml
import shutil
import random


# the path of the dataset
data_path = '/home/carey/Projects/GoldenAssistant/SystemCode/core/yolo/heroes/dataset'
train_ratio = 0.8
val_ratio = 0.1
test_ratio = 0.1
raw_data_path = '.'
dataset_augment = True
raw_image_path = os.path.join(raw_data_path, 'images')
raw_label_path = os.path.join(raw_data_path, 'labels')

new_image_path = os.path.join(data_path, 'images')
new_label_path = os.path.join(data_path, 'labels')
new_image_train_path = os.path.join(data_path, 'images/train')
new_label_train_path = os.path.join(data_path, 'labels/train')
new_image_val_path = os.path.join(data_path, 'images/val')
new_label_val_path = os.path.join(data_path, 'labels/val')
new_image_test_path = os.path.join(data_path, 'images/test')
new_label_test_path = os.path.join(data_path, 'labels/test')

# create the new folders
if os.path.exists(data_path):
    shutil.rmtree(data_path)

os.makedirs(new_image_path, exist_ok=True)
os.makedirs(new_label_path, exist_ok=True)
os.makedirs(new_image_train_path, exist_ok=True)
os.makedirs(new_label_train_path, exist_ok=True)
os.makedirs(new_image_val_path, exist_ok=True)
os.makedirs(new_label_val_path, exist_ok=True)
os.makedirs(new_image_test_path, exist_ok=True)
os.makedirs(new_label_test_path, exist_ok=True)

# get all the image files
image_files = [os.path.join(raw_image_path, f) for f in os.listdir(raw_image_path) if f.endswith('.jpg')]
# get all the label files
label_files = [os.path.join(raw_label_path, f) for f in os.listdir(raw_label_path) if f.endswith('.txt')]
if dataset_augment:
    augment_images = [(os.path.join('images_augmented', 'images', f), os.path.join('images_augmented', 'labels', f.replace('.jpg', '.txt'))) for f in os.listdir(os.path.join('images_augmented', 'images')) if f.endswith('.jpg')]
    # augment_labels = [os.path.join('images_augmented', 'labels', f) for f in os.listdir(os.path.join('images_augmented', 'labels')) if f.endswith('.txt')]
    image_files += augment_images

# get the total number of the dataset
total_num = len(image_files)
# get the number of the train, val, test dataset
train_num = int(total_num * train_ratio)
val_num = int(total_num * val_ratio)
test_num = total_num - train_num - val_num

# train_files
train_images = random.sample(image_files, k=train_num)
val_images = random.sample(list(set(image_files) - set(train_images)), k=val_num)
test_images = list(set(image_files) - set(train_images) - set(val_images))

# copy the train images and labels
for image in train_images:
    if type(image) == tuple:
        shutil.copy(image[0], new_image_train_path)
        shutil.copy(image[1], new_label_train_path)
        continue

    shutil.copy(image, new_image_train_path)
    shutil.copy(image.replace('.jpg', '.txt').replace('images', 'labels'), new_label_train_path)

# copy the val images and labels
for image in val_images:
    if type(image) == tuple:
        shutil.copy(image[0], new_image_val_path)
        shutil.copy(image[1], new_label_val_path)
        continue

    shutil.copy(image, new_image_val_path)
    shutil.copy(image.replace('.jpg', '.txt').replace('images', 'labels'), new_label_val_path)

# copy the test images and labels
for image in test_images:
    if type(image) == tuple:
        shutil.copy(image[0], new_image_test_path)
        shutil.copy(image[1], new_label_test_path)
        continue

    shutil.copy(image, new_image_test_path)
    shutil.copy(image.replace('.jpg', '.txt').replace('images', 'labels'), new_label_test_path)

# load names
with open('datasets.yaml', 'r') as f:
    label_info = yaml.load(f, Loader=yaml.FullLoader)

# generate yolo format yaml file
data = dict(
    path=f'{os.path.abspath(data_path)}',
    train='images/train',
    val=f'images/val',
    test=f'images/test',
    names=label_info['names']
)

with open(f'{data_path}/dataset.yaml', 'w') as f:
    yaml.dump(data, f, allow_unicode=True)