Automatic Image captioning is one of the most recent problem to attract both Computer Vision and Natural Language community. Object and pattern recognition in images, as well as relationships between words, has to be trained in order to successfully come up with an image caption.
In this project, I would attempt to use a pre-trained Convolutional Neural Network (Xception) to process image data, followed by building a deep multi-layer recurrent neural network to generate the captions in English.