create_input_files_f30k.py

from utils import create_input_files

"""
To create files that contain all images stored in h5py format and captions stored in json files.
Minimum word frequencies to be used as cut-off for removing rare words to be specified here.
"""

if __name__ == '__main__':
    create_input_files(dataset='flickr30k',
                       karpathy_json_path='path_to___dataset_flickr30k.json',
                       image_folder='path_to_flickr30k_images_folder',
                       captions_per_image=5,
                       min_word_freq=5,
                       output_folder='folder_for_processed_data',
                       max_len=70)

# 76 is the maximum length of sentences in Flickr30k dataset