CASTLE 2024 Dataset
收藏资源简介:
CASTLE 2024数据集是一个包含第一人称和第三人称视角的多模态视频数据集,由都柏林城市大学等机构的研究人员创建。该数据集包含了10名参与者在4天内进行的一系列日常生活活动,总计超过600小时的超高清视频资料。它提供了15个时间对齐的视频源,包括10个参与者佩戴的GoPro HERO10相机和5个固定摄像头。数据集记录了参与者的真实互动,未进行面部模糊或音频扭曲等匿名处理,以保持数据的丰富性和真实性。该数据集适用于多媒体分析理解任务,如生活日志检索、对象和动作识别、社交互动分析以及动态3D场景的重建和分析。
The CASTLE 2024 dataset is a multimodal video dataset featuring first-person and third-person perspectives, created by researchers from institutions including Dublin City University. It comprises a series of daily living activities conducted by 10 participants over 4 days, with a total of over 600 hours of ultra-high-definition video footage. The dataset provides 15 temporally aligned video sources, including 10 GoPro HERO10 cameras worn by the participants and 5 fixed cameras. It records the real-world interactions of the participants, with no anonymization measures such as facial blurring or audio distortion applied, to preserve the richness and authenticity of the data. This dataset is applicable to multimedia analysis and understanding tasks, such as life log retrieval, object and action recognition, social interaction analysis, and dynamic 3D scene reconstruction and analysis.




