Scene Processing With Machine Learnable and Semantically Parametrized Representations RENEWAL

使用机器学习和语义参数化表示进行场景处理 RENEWAL

• 批准号: MR/Y033884/1
• 负责人: Ahmet Oztireli
• 金额: $ 75.36万
• 依托单位: University of Cambridge
• 依托单位国家: 英国
• 项目类别: Fellowship
• 财政年份: 2025
• 资助国家: 英国
• 起止时间: 2025 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=MR%2FY033884%2F1
• 关键词: Scene; Processing; Machine; Learnable; Semantically; Scene; Processing; Machine; Learnable; Semantically

Generative artificial intelligence has made a significant leap in developing intuitive natural language based computer interfaces via large language models, and realistic image and video generation from text inputs. These have already started revolutionising creation and editing of code, text, images, videos, presentations, and many other digital media. We will extend these techniques to scene creation and processing by relying on the techniques we have been developing over the years. We will develop text-to-scene models for easy creation, intuitive control, and conversational capture of 3D objects and scenes. We will deploy these models to the new generation of extended reality devices with advanced displays, optics, sensors, and processing units to build a scene creation and editing system that uses voice and brain signals as inputs.

生成人工智能通过大型语言模型以及文本输入中的现实图像和视频生成开发直觉的自然计算机界面方面取得了重大飞跃。这些已经开始改变代码，文本，图像，视频，演示文稿和许多其他数字媒体的创建和编辑。我们将通过依靠多年来发展的技术来将这些技术扩展到场景创建和处理。我们将开发文本对场景模型，以方便创建，直观控制以及3D对象和场景的对话捕获。我们将这些模型将这些模型部署到新一代的扩展现实设备，其中具有高级显示器，光学，传感器和处理单元，以构建场景创建和编辑系统，该系统使用语音和大脑信号作为输入。