PPS-D: Person-oriented Protection of Public Spaces Dataset

PPS-D is a comprehensive synthetic image and video dataset designed explicitly for the protection of public spaces. This dataset is compiled from a large number of video surveillance cameras that monitor different environments such as streets, squares, parks, railway stations, airports, car parks, shopping centres and industrial areas, as well as different weather and time conditions such as sun, clouds, night, rain, snow, etc. Each camera is strategically positioned to provide a realistic view of the most important areas in each scenario. All the environments and actors, including people, vehicles and objects, are entirely synthetic and created from scratch. The generated individuals are not based on any real or specific person.PPS-D emphasises object and event detection, with various simulations representing different events, all linked to human behaviour. The dataset includes scenarios ranging from everyday normality to more complex situations, such as abandoned objects, increasing crowd density and sudden panic caused by pedestrian or vehicle incidents. The diverse settings and realistic camera perspectives are designed to enhance algorithms for public safety and surveillance applicationsPPS-D是专为公共空间安全防护打造的综合性合成图像与视频数据集。该数据集源自大量部署于不同场景的视频监控摄像头，覆盖街道、广场、公园、火车站、机场、停车场、购物中心及工业区等各类环境，同时涵盖晴天、多云、夜间、降雨、降雪等不同天气与时段条件。每台摄像头均经过策略性部署，可获取对应场景中核心区域的真实视角。所有场景与场景内的主体（包括行人、车辆与各类物体）均为完全合成生成，从零构建，其中生成的行人个体未基于任何真实或特定自然人。PPS-D聚焦目标与事件检测任务，包含多种模拟生成的不同事件场景，所有事件均与人类行为相关联。该数据集涵盖的场景覆盖日常常态至复杂异常场景，例如遗留物品、人群密度攀升以及由行人或车辆事故引发的突发恐慌等。多样化的场景设置与真实的摄像头视角，旨在优化面向公共安全与监控应用的算法性能。