MovieQA 2017 Челлендж

AbstractWe propose a new model for speaker naming in movies that leverages visual, textual, and acoustic modalities in an unified optimization framework. To evaluate the performance of our model, we …...