Evaluating visual "common sense" using fine-grained classification and captioning tasksDownload PDF

12 Feb 2018, 19:56 (modified: 04 Jun 2018, 15:00)ICLR 2018 Workshop SubmissionReaders: Everyone
Keywords: video dataset, fine-grained understanding, common sense, video captioning
Abstract: We introduce the Something-something V2 dataset, which contains captions of finely-varying human-object interactions. We also discuss various baseline models, and show that neural networks show surprisingly strong performance on many of the very hard, detailed discrimination tasks associated with this dataset.
4 Replies