Retrained the model using Flickr 30k image dataset which contains 158873 image caption.
Edited the generate function to redirect the output to a file (optional). Pass return_as_list=True to the generate function to get the output as list (preferred).
Requirements: textgenrnn that uses Python3 and run on top on Tensorflow
Attached: Flickr 30k captions In.txt and the trained model textgenrnn_weights.hdf5
Examples:
Two teenagers are sitting on a sidewalk .
A man is sitting on a bench looking at a park .
Two people are standing on a rock .
Man in a suit is in a field .
Two young man in black shirt and a worker shorts playing a soccer ball .
Two children are walking in front of a statue .
References:
Original Github repo
Flickr