Task-Driven Dynamic Fusion: Reducing Ambiguity in Video Description | IEEE Conference Publication | IEEE Xplore