Top-view videos

The top-down video dataset is created using the downward pointing camera of the simulator program. This data set is generated using the Coppelia Sim video recording feature. The captured videos are coloured frames of size 720x1280. These videos are converted to grey scale (single channel). Then they are compressed to 240x320 resolution to facilitate the proposed computations in a feasible time. The final size of the dataset is 493, 5-second (75 frame) videos for structured and random behaviours.

本俯视视角视频数据集通过仿真程序的朝下拍摄摄像头采集构建，依托Coppelia Sim的视频录制功能生成。采集得到的原始视频为分辨率720×1280的彩色帧序列，随后被转换为灰度（单通道）格式，再压缩至240×320分辨率，以确保本文所提出的计算任务可在合理耗时内完成。本数据集最终包含493段时长为5秒（含75帧）的视频，涵盖结构化行为与随机行为两类场景。