import os from kaggle.api.kaggle_api_extended import KaggleApi import shutil import tarfile data_dir = 'data/' kaggle_api = KaggleApi() kaggle_api.authenticate() kaggle_api.dataset_download_files('gpiosenka/100-bird-species', path=data_dir, unzip=True) # There is a bug in the dataset, where one of the folders is named "PARAKETT AUKLET" instead of "PARAKETT AUKLET" # We fix it by adding a space to the valid folder, because train and test are wrong and also the names in the labels file # Fixing the path fault_path = os.path.join(data_dir, 'valid', 'PARAKETT AUKLET') correct_path = os.path.join(data_dir, 'valid', 'PARAKETT AUKLET') shutil.rmtree(correct_path, ignore_errors=True) shutil.move(fault_path, correct_path) # Compressing the train directory with tarfile.open(os.path.join(data_dir, 'train.tar.gz'), 'w:gz') as tar: tar.add(os.path.join(data_dir, 'train'), arcname=os.path.basename(os.path.join(data_dir, 'train'))) # Compressing the test directory with tarfile.open(os.path.join(data_dir, 'test.tar.gz'), 'w:gz') as tar: tar.add(os.path.join(data_dir, 'test'), arcname=os.path.basename(os.path.join(data_dir, 'test'))) # Compressing the valid directory with tarfile.open(os.path.join(data_dir, 'valid.tar.gz'), 'w:gz') as tar: tar.add(os.path.join(data_dir, 'valid'), arcname=os.path.basename(os.path.join(data_dir, 'valid'))) os.remove(os.path.join(data_dir, 'EfficientNetB0-525-(224 X 224)- 98.97.h5'))