ViGiL3D
收藏资源简介:
ViGiL3D是由西蒙弗雷泽大学和阿尔伯塔机器智能研究所创建的一个3D视觉定位数据集,旨在通过多样化的语言模式来评估视觉定位方法的性能。该数据集包含350条数据,涵盖了多种语言现象,如否定、指代解析等。数据集的内容包括3D场景和自然语言描述,数据来源为手动标注。ViGiL3D的应用领域包括计算机图形学、机器人学以及虚拟和增强现实助手的对话系统,旨在解决现有3D视觉定位模型在处理多样化语言模式时的不足。
ViGiL3D is a 3D visual grounding dataset developed by Simon Fraser University and the Alberta Machine Intelligence Institute, which aims to evaluate the performance of visual grounding approaches through diverse linguistic patterns. This dataset consists of 350 samples, covering a wide range of linguistic phenomena including negation and reference resolution. Its content comprises 3D scenes and natural language descriptions, with all data manually annotated. The applicable fields of ViGiL3D cover computer graphics, robotics, and dialogue systems for virtual and augmented reality assistants, and it is designed to address the shortcomings of existing 3D visual grounding models when handling diverse linguistic patterns.




