Image Caption Generator

A Neural Network based generative model for captioning images.

##Pre-Requisites:

Tensorflow
NLTK
pandas
skimage
Download Flickr30K Dataset: Images and results_20130124.token
Download Pre-Trained InceptionV4 Tensorflow graph from DeepDetect available here

##Procedure to Run:

Clone the Repository to preserve Directory Structure
Put results_20130124.token and Flickr30K images in Dataset Folder.
Put inception_v4.pb in CNNs folder
Run train.py with parameter resume = 0, to start fresh training procedure. resume = 1 restores training process from latest saved model.
Run test.py with path=IMAGE_PATH/URL to generate captions OR follow test.ipynb

##Reference: Show and Tell: A Neural Image Caption Generator

-Oriol Vinyals, Alexander Toshev, Samy Bengio, Dumitru Erhan

##License: Protected Under BSD-3 Clause License.

Name		Name	Last commit message	Last commit date
Latest commit History 35 Commits
CNNs		CNNs
Dataset		Dataset
ImgCap		ImgCap
References		References
model		model
CNN.py		CNN.py
LICENSE		LICENSE
README.md		README.md
caption.py		caption.py
split_data.py		split_data.py
test.ipynb		test.ipynb
test.py		test.py
train.py		train.py

Provide feedback