This repository (under construction) contains the code for the CoNLL 2021 submission VQA-MHUG: A gaze dataset to study multimodal neural attention in VQA