Automatic Neural Image Captioning

I am using an inject architecture for Xception - LSTM model to generate caption for images. I used pretrained weights from ImageNet for Xception encoder and used pretrained Stanford GloVe word embeddings.

Data

I used Flickr 8k dataset which is a collection of 8k images with their respective captions where each image consists five different captions.

Requirements

Python 3.7

Keras 2.3

Numpy 1.18

Pandas 1.0

Matplotlib 3.1

NLTK 3.4.5

Re 2.2.1

Results

Reference

Deep Visual-Semantic Alignments for Generating Image Descriptions, Andrej Karpathy, Li Fei-Fei