arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2008.09849cs.CV
收藏

视频问答任务的数据增强技术

Data augmentation techniques for the Video Question Answering task

Fondazione Bruno Kessler(布鲁诺·凯斯勒基金会);University of Udine(乌迪内大学)

Alex Falcon, Oswald Lanz, Giuseppe Serra

摘要:

视频问答(VideoQA)是一项需要模型来分析和理解输入视频给出的视觉内容和问题给出的文本部分以及它们之间的交互以产生有意义的答案的任务。在我们的工作中,我们专注于以自我为中心的视频问答任务,它利用第一人称视频,因为这种任务的重要性,它可以对许多不同的领域产生影响,例如与社会援助和工业培训有关的领域。最近,一个以自我为中心的视频QA数据集,称为EgoVQA,已经发布。由于体积小,模型往往很快就会过拟合。为了缓解这个问题,我们提出了几种增强技术,使我们在考虑的基线上的最终精度提高了+5.5%。

英文摘要:

Video Question Answering (VideoQA) is a task that requires a model to analyze and understand both the visual content given by the input video and the textual part given by the question, and the interaction between them in order to produce a meaningful answer. In our work we focus on the Egocentric VideoQA task, which exploits first-person videos, because of the importance of such task which can have impact on many different fields, such as those pertaining the social assistance and the industrial training. Recently, an Egocentric VideoQA dataset, called EgoVQA, has been released. Given its small size, models tend to overfit quickly. To alleviate this problem, we propose several augmentation techniques which give us a +5.5% improvement on the final accuracy over the considered baseline.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑