mehmetkeremturkcan/bikeped

bikeped一致性场景数据集包含24个脚本化的行人与骑行者交叉路口场景，用于评估[bikeped](https://github.com/mkturkcan/bikeped)中的碰撞预警系统。每个场景定义了一些代理（行人、骑行者、电动自行车）在地平面上通过手工编写的路径点序列移动；测试平台以30帧每秒的速度对这些路径点进行插值，并从保守的运动学安全模型中分配每帧的真实危险标签。该数据集是技术报告《A Real-Time Bike–Pedestrian Safety System with Wide-Angle Perception and Evaluation Testbed for Urban Intersections》的配套数据。数据集结构包括两种配置：`scenarios`（包含24行，每行一个场景的原始路径点路径）和`frames`（包含12901行，每行一个`(scenario, agent, frame)`的插值地面位置和危险评估）。数据集还包括两种视频渲染（示意图和照片级真实感渲染）以及详细的运动学真实模型用于危险标注。

The bikeped Conformance Scenarios dataset consists of 24 scripted pedestrian–cyclist intersection scenarios used to evaluate the collision-warning system in [bikeped](https://github.com/mkturkcan/bikeped). Each scenario defines a handful of agents (pedestrians, cyclists, e-bikes) moving on the ground plane through a hand-authored sequence of waypoints; the testbench interpolates these at 30 fps and assigns per-frame ground-truth danger labels from a conservative kinematic safety model. Companion to the technical report *A Real-Time Bike–Pedestrian Safety System with Wide-Angle Perception and Evaluation Testbed for Urban Intersections*. The dataset structure includes two configs: `scenarios` (24 rows with raw waypoint paths) and `frames` (12901 rows with denormalized data per `(scenario, agent, frame)`). It also includes two video renderings (schematic and photorealistic) and a detailed kinematic ground-truth model for danger assessment.