BLVD

该数据集名为BLVD，包含了654个来自江苏省常熟市的高分辨率视频片段，总计120,000帧。它记录了6,004个有效的事件片段，包括周围参与者的记录ID、3D坐标、方向信息以及所有物体的交互行为。此外，该数据集根据不同的光照条件被划分为训练集和测试集，具体分为白天高密度、白天低密度、夜间高密度和夜间低密度。规模上，数据集包含了654个视频片段、120,000帧和6,004个有效事件片段。其任务是对自动驾驶车辆周围的目标代理进行轨迹预测。

This dataset, named BLVD, consists of 654 high-resolution video clips sourced from Changshu City, Jiangsu Province, totaling 120,000 frames. It documents 6,004 valid event segments, including the recording IDs of surrounding participants, 3D coordinates, orientation information, and interactive behaviors of all objects involved. Additionally, this dataset is partitioned into training and test sets based on varying lighting conditions, with four specific categories: high-density daytime, low-density daytime, high-density nighttime, and low-density nighttime. In terms of overall scale, the dataset contains 654 video clips, 120,000 frames, and 6,004 valid event segments. The primary task of this dataset is trajectory prediction for target agents in the vicinity of autonomous vehicles.