1 paper
Sebastian Kmiec, Juhan Bae, Ruijian An
We introduce modifications to state-of-the-art approaches to aggregating local video descriptors by using attention mechanisms and function approximations. Rather than using ensemb…