arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.16565cs.CV
收藏

学习表示不变性以实现数据高效的动作识别

Learning Representational Invariances for Data-Efficient Action Recognition

Virginia Tech(弗吉尼亚理工大学);Kyung Hee University(庆熙大学);University of Delaware(特拉华大学);University of Maryland College Park(马里兰大学帕克分校)

Yuliang Zou, Jinwoo Choi, Qitong Wang, Jia-Bin Huang

摘要:

数据增强是一种普遍存在的技术,用于改善图像分类时,标记的数据是稀缺的。将模型预测约束为对不同的数据增强不变有效地将期望的表示不变性注入到模型(例如,不变性光度变化),并有助于提高精度。与图像数据相比,由于额外的时间维度,视频中的外观变化要复杂得多。然而,用于视频的数据增强方法仍然未被充分探索。本文研究了各种数据增强策略,捕捉不同的视频不变性,包括光度,几何,时间,演员/场景增强。当与现有的半监督学习框架相结合时,我们证明了我们的数据增强策略在低标签制度下对Kinetics-100/400,Mini-Something-v2,UCF-101和HMDB-51数据集产生了有希望的性能。我们还验证了我们的数据增强策略,在完全监督的设置,并表现出改善的性能。

英文摘要:

Data augmentation is a ubiquitous technique for improving image classification when labeled data is scarce. Constraining the model predictions to be invariant to diverse data augmentations effectively injects the desired representational invariances to the model (e.g., invariance to photometric variations) and helps improve accuracy. Compared to image data, the appearance variations in videos are far more complex due to the additional temporal dimension. Yet, data augmentation methods for videos remain under-explored. This paper investigates various data augmentation strategies that capture different video invariances, including photometric, geometric, temporal, and actor/scene augmentations. When integrated with existing semi-supervised learning frameworks, we show that our data augmentation strategy leads to promising performance on the Kinetics-100/400, Mini-Something-v2, UCF-101, and HMDB-51 datasets in the low-label regime. We also validate our data augmentation strategy in the fully supervised setting and demonstrate improved performance.

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。

↑