TITAN

该数据集名为TITAN，包含了10小时的驾驶视频，以每秒60帧的速度录制，并针对轨迹预测和多级别动作识别，标注了700个短片段。此外，该数据集提供了按情境复杂度分类的各种动作标签，重点关注行人的行为。规模上，数据集涵盖了10小时的视频，共有8,592个独特的行人。其任务旨在进行轨迹预测和动作识别。

This dataset is named TITAN. It contains 10 hours of driving video recorded at 60 frames per second, and is annotated with 700 short segments for trajectory prediction and multi-level action recognition. Additionally, the dataset provides various action labels categorized by scenario complexity, with a primary focus on pedestrian behaviors. In terms of scale, the dataset spans 10 hours of video and includes 8,592 unique pedestrians. The core tasks of this dataset are trajectory prediction and action recognition.