arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2108.09668cs.CV
收藏

学习视觉关系:魔鬼在尾巴里

Learning of Visual Relations: The Devil is in the Tails

University of California San Diego(加州大学圣迭戈分校);Intel Labs(英特尔实验室)

Alakh Desai, Tz-Ying Wu, Subarna Tripathi, Nuno Vasconcelos

摘要:

最近,大量的工作致力于建模视觉关系。这主要解决了架构的设计,通常是通过添加参数和增加模型复杂性。然而,视觉关系学习是一个长尾问题,由于联合推理的组合性质的群体的对象。一般来说,增加模型复杂性不适合长尾问题,因为它们倾向于过拟合。在本文中,我们探讨了另一种假设,表示魔鬼在尾巴。在此假设下,通过保持模型简单但提高其处理长尾分布的能力来实现更好的性能。为了验证这一假设,我们设计了一种新的方法来训练视觉关系模型,这是受到最先进的长尾识别文献的启发。这是基于迭代解耦训练方案,表示为尾中魔鬼的解耦训练(DT 2)。DT 2采用了一种新的抽样方法,交替类平衡抽样(ACBS),捕捉视觉关系的长尾实体和谓词分布之间的相互作用。结果表明,一个非常简单的架构,DT 2-ACBS显着优于更复杂的场景图生成任务的最先进的方法。这表明,必须考虑与问题的长尾性质同步发展复杂的模型。

英文摘要:

Significant effort has been recently devoted to modeling visual relations. This has mostly addressed the design of architectures, typically by adding parameters and increasing model complexity. However, visual relation learning is a long-tailed problem, due to the combinatorial nature of joint reasoning about groups of objects. Increasing model complexity is, in general, ill-suited for long-tailed problems due to their tendency to overfit. In this paper, we explore an alternative hypothesis, denoted the Devil is in the Tails. Under this hypothesis, better performance is achieved by keeping the model simple but improving its ability to cope with long-tailed distributions. To test this hypothesis, we devise a new approach for training visual relationships models, which is inspired by state-of-the-art long-tailed recognition literature. This is based on an iterative decoupled training scheme, denoted Decoupled Training for Devil in the Tails (DT2). DT2 employs a novel sampling approach, Alternating Class-Balanced Sampling (ACBS), to capture the interplay between the long-tailed entity and predicate distributions of visual relations. Results show that, with an extremely simple architecture, DT2-ACBS significantly outperforms much more complex state-of-the-art methods on scene graph generation tasks. This suggests that the development of sophisticated models must be considered in tandem with the long-tailed nature of the problem.

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。

↑