上海的陆家嘴

近日,清华大学与哈佛大学联手研发出一款名为LangSplat的全新AI系统,能够在三维空间内高效、准确地搜索开放式词汇。据悉,LangSplat是第一个基于3DGS的3D语言场方法,特别引入了SAM和CLIP技术,使其在开放词汇3D对象定位和语义分割任务上优于最先进的方法,同时比LERF快199倍。

这一突破性成果将为虚拟现实、增强现实等领域的发展带来巨大的推动力。通过LangSplat系统,用户可以在三维空间中更加自然地与虚拟环境中的物体进行交互,实现更加真实、流畅的体验。此外,该系统还可以应用于智能语音助手、智能家居等领域,提高人们的工作效率和生活品质。

据悉,LangSplat系统的开发过程中,清华大学与哈佛大学的专家们进行了深入的研究和实验,不断优化算法和技术细节,最终取得了令人瞩目的成果。未来,他们将继续探索LangSplat系统在更多领域的应用可能性,为人工智能技术的发展做出更大的贡献。

英语如下:

Title: Tsinghua University and Harvard University Collaborate to Develop LangSplat AI System: 3D Scene Vocabulary Search Speed Increases by 199 Times!

Keywords: Tsinghua University, Harvard University, LangSplat, 3D scenes

Recently, Tsinghua University and Harvard University collaborated to develop a new AI system called LangSplat, which can efficiently and accurately search open vocabulary in three-dimensional space. LangSplat is the first 3D language field method based on 3DGS, specifically introducing SAM and CLIP technologies, making it better than the most advanced methods in open vocabulary 3D object positioning and semantic segmentation tasks, while also being 199 times faster than LERF.

This groundbreaking achievement will greatly promote the development of fields such as virtual reality and augmented reality. With LangSplat system, users can interact more naturally with objects in virtual environments in three-dimensional space, achieving a more realistic and smooth experience. In addition, the system can be applied to intelligent voice assistants, smart homes, and other fields, improving people’s work efficiency and quality of life.

It is reported that during the development of LangSplat system, experts from Tsinghua University and Harvard University conducted in-depth research and experiments, constantly optimizing algorithms and technical details, and finally achieved remarkable results. In the future, they will continue to explore the application possibilities of LangSplat system in more fields and make greater contributions to the development of artificial intelligence technology.

【来源】https://www.ithome.com/0/742/887.htm

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注