arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2107.07201cs.CV
收藏

用于视觉和语言导航的邻近视图增强模型

Neighbor-view Enhanced Model for Vision and Language Navigation

Center for Research on Intelligent Perception and Computing, Institution of Automation, Chinese Academy of Sciences(中国科学院自动化研究所智能感知与计算研究中心);School of Future Technology, University of Chinese Academy of Sciences(中国科学院大学未来技术学院);University of Adelaide(阿德莱德大学);Center for Excellence in Brain Science and Intelligence Technology (CEBSIT)(脑科学与智能技术卓越创新中心(CEBSIT));Chinese Academy of Sciences, Artificial Intelligence Research (CAS-AIR)(中国科学院人工智能研究院(CAS-AIR))

Dong An, Yuankai Qi, Yan Huang, Qi Wu, Liang Wang, Tieniu Tan

摘要:

视觉和语言导航(VLN)要求智能体通过遵循自然语言指令导航到目标位置。现有的研究大多是通过候选导航对象所在的单个视图的特征来表示候选导航对象。然而,指令可能提及单个视图之外的地标作为参考,这可能导致现有方法的文本-视觉匹配失败。在这项工作中,我们提出了一个多模块的邻居视图增强模型(NvEM),自适应地将视觉上下文从邻居视图更好的纹理视觉匹配。具体来说,我们的NvEM利用一个主题模块和一个参考模块来收集来自邻居视图的上下文。主题模块在全局级别融合相邻视图,并且参考模块在局部级别融合相邻对象。主语和指称是通过注意机制自适应地确定的。我们的模型还包括一个动作模块,以利用强方向指导(例如,“向左拐”)。每个模块分别预测导航动作,并将其加权和用于预测最终动作。大量的实验结果表明,所提出的方法对R2 R和R4 R基准对几个国家的最先进的导航仪的有效性,和NvEM甚至击败一些预训练的。我们的代码可在https://github.com/MarSaKi/NvEM上获得。

英文摘要:

Vision and Language Navigation (VLN) requires an agent to navigate to a target location by following natural language instructions. Most of existing works represent a navigation candidate by the feature of the corresponding single view where the candidate lies in. However, an instruction may mention landmarks out of the single view as references, which might lead to failures of textual-visual matching of existing methods. In this work, we propose a multi-module Neighbor-View Enhanced Model (NvEM) to adaptively incorporate visual contexts from neighbor views for better textual-visual matching. Specifically, our NvEM utilizes a subject module and a reference module to collect contexts from neighbor views. The subject module fuses neighbor views at a global level, and the reference module fuses neighbor objects at a local level. Subjects and references are adaptively determined via attention me'chanisms. Our model also includes an action module to utilize the strong orientation guidance (e.g., "turn left") in instructions. Each module predicts navigation action separately and their weighted sum is used for predicting the final action. Extensive experimental results demonstrate the effectiveness of the proposed method on the R2R and R4R benchmarks against several state-of-the-art navigators, and NvEM even beats some pre-training ones. Our code is available at https://github.com/MarSaKi/NvEM.

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。

↑