distribute images to right sub folder for training

distributor.py

import pandas as pd
import os

train = pd.read_csv('train.csv')
landmarks = list(set(train['landmark_id'].tolist())) #use set to have only unique id's and turn it back to a list type again

#creates sub folders of landmark id in train and validation folder
train_folder = 'training_images'
validation_folder = 'validation_images'

for landmark in landmarks:
    train_dir = "{0}/{1}".format(train_folder, landmark)
    val_dir = "{0}/{1}".format(validation_folder, landmark)
  
    if not os.path.exists(train_dir):
        os.makedirs(train_dir)
    
    if not os.path.exists(val_dir):
        os.makedirs(val_dir)
        
#now you need to match all the images downloaded to training_images to the right subfolder

#create image names and get landmarks
images_ids = ["{0}.jpg".format(x) for x in train['id'].tolist()]
full_landmarks = train['landmark_id'].tolist()

for image_name, landmark in zip(images_ids, full_landmarks):
    old_filename = "{0}/{1}".format(train_folder, image_name)
    new_filename = "{0}/{1}/{2}".format(train_folder, landmark, imagename)
    try:
        os.rename(old_filename, new_filename)
    except Exception as e:
        print ( "Failed to move file {0} error {1}".format(old_filename, e) )
        
#move 20% of the images to the validation folder
for folder in landmarks:
    files = [f for f in os.listdir('{0}/{1}'.format(train_folder,folder)) if os.path.isfile(join('{0}/{1}'.format(train_folder, folder), f))] 
    sample_files = files[:len(files)/5]
    
    for image in small_slice:
        try:
            os.rename('{0}/{1}/{2}'.format(train_folder, folder, image), '{0}/{1}/{2}'.format(validation_folder, folder, image))
        except Exception as e:
            print "Failed to move file {0} error {1}".format(image, e)