TVQA+: Spatio-Temporal Grounding for Video Question Answering
2019·Arxiv