urbanaudiosensing/ASPEDvb

ASPED v.b数据集是一个全面的、包含1,321小时路边音频和视频录音的集合，旨在用于在有车辆噪声的环境中进行行人检测。该数据集在乔治亚理工学院校园内的单一位置（第五街）通过多个摄像头和录音设备设置收集而成，包含了4个不同时间段的录音。每个录音都包括16 kHz单声道音频，与帧级行人注释同步，以及每秒1帧的视频缩略图。该数据集主要用于音频行人检测任务，也可用于噪声环境中的声音事件检测、声学模型的领域适应和城市声景分析等相关任务。

The ASPED v.b dataset is a comprehensive collection of 1,321 hours of roadside audio and video recordings designed for pedestrian detection in the presence of vehicular noise. The dataset was collected from multiple camera and recorder setups at a single location (Fifth Street) on the Georgia Institute of Technology campus and contains recordings from 4 different sessions. Each recording includes 16 kHz mono audio synchronized with frame-level pedestrian annotations and 1 fps video thumbnails. The dataset is primarily intended for audio-based pedestrian detection and can also be used for related tasks such as sound event detection in noisy environments, domain adaptation for acoustic models, and urban soundscape analysis.