Loads the dataset in batches for training :return: batch of data

:param target_seq: the array that is fed into the model to predict the next word :param states_value: previous state that is fed int

:param target_seq: the array that is fed into the model to predict the next word :param states_value: previous state

saves the numpy features from all the videos

:param video: The video whose frames are to be extracted to convert into a numpy array :param model: the pretrained vgg16 model :return:

loads all the numpy files :return: two lists containing all the video arrays and the video Id

:param f: the loaded numpy array after creating videos to frames and extracting features :return: the final sentence which has been

:param f: the loaded numpy array after creating videos to frames and extracting features :return: the final sentence

:param filename: the video to load :param caption: final caption :return:

Preprocessing the data dumps values of the json file into a list

writes the captions of all the testing videos in a text file

an encoder decoder sequence to sequence model reference : https://arxiv.org/abs/1505.00487

Returns the model that will be used for inference