Cross-Sentence Temporal and Semantic Relations in Video Activity Localisation----ICCV2021论文阅读Video-Sentence AlignmentCross-Sentence Relations MiningModel Training
目录Video-Sentence AlignmentMatching Score.Multi-Instance Learning.Cross-Sentence Relations MiningTemporal Consistency.Semantic Consistency.Model Training这篇论文的Task是视频定位只有视频文本的对应关系,但是没有ground truth的时间边界,因此是弱监督的现有的弱监督解决方案首先分别定位不同的MoIs,但这不是最