Learning Visually Grounded Sentence Representations

Douwe Kiela, Alexis Conneau, Allan Jabri +1

We introduce a variety of models, trained on a supervised image captioning corpus to predict the image features for a given caption, to perform sentence representation grounding. W…