YouTube-8M dataset file renaming for avoiding duplicate file names in Windows (Python code)

## To avoid duplicate file names problem in Windows (e.g. Ab.txt, aB.txt),

## At first, I download the dataset and renamed the file names in the Linux OS.

## then, I copy renamed files to my Windows desktop for training deep neural network.

## Renamed dataset also works well on the YouTube-8M training code.

import os

import shutil

oldpath = '/data/yt8m/v2/video/'

newpath = '/data/yt8m/v2/video2/'

file_list = os.listdir(oldpath)

file_train_cnt = 0

file_test_cnt = 0

file_validate_cnt = 0

new_file_name = ''

for filename in file_list:

old_file_path = os.path.join(oldpath, filename)

if 'tra' in filename[:3] and '.tfrecord' in filename:

new_file_name = 'train' + str(file_train_cnt).zfill(5) + '_' + \

str(filename).split('.')[0] + '.tfrecord'

file_train_cnt += 1

elif 'tes' in filename[:3] and '.tfrecord' in filename:

new_file_name = 'test' + str(file_test_cnt).zfill(5) + '_' + \

str(filename).split('.')[0] + '.tfrecord'

file_test_cnt += 1

elif 'val' in filename[:3] and '.tfrecord' in filename:

new_file_name = 'validate' + str(file_validate_cnt).zfill(5) + \

'_' + str(filename).split('.')[0] + '.tfrecord'

file_validate_cnt += 1

else:

continue

new_file_path = os.path.join(newpath, new_file_name)

with open(old_file_path, 'rb') as f1, open(new_file_path, 'wb') as f2:

f2.write(f1.read())

print('Successfuly Copied! [', filename, ' to ', new_file_path, ']')

print(file_train_cnt)

print(file_test_cnt)

print(file_validate_cnt)