traffic-vehicle-detection
收藏资源简介:
该数据集是Edge-AI交通车辆检测数据集(UA-DETRAC CCTV),源自Edge-AI交通与车辆分析系统项目。它基于UA-DETRAC基准,从固定交叉口监控摄像头中整理并标准化了23,319张CCTV交通图像,共包含215,109个标注边界框,采用标准YOLO格式,涵盖4类车辆:car(小汽车,177,403个边界框,占82.5%)、bus(公交车,3,523个,1.6%)、truck(卡车,16,051个,7.5%)和van(面包车,18,132个,8.4%)。数据集已划分为训练集(18,655张图像,占80%)、验证集(2,332张,10%)和测试集(2,332张,10%)。所有图像分辨率归一化为640x640(原生分辨率为960x540)。该数据集适用于目标检测任务,特别是车辆检测场景,可用于智能交通、智慧城市等应用。许可证为CC-BY-4.0。
This dataset is the Edge-AI Traffic Vehicle Detection dataset (UA-DETRAC CCTV), derived from the Edge-AI Traffic and Vehicle Analysis System project. Based on the UA-DETRAC benchmark, it compiles and standardizes 23,319 CCTV traffic images from fixed intersection surveillance cameras, containing a total of 215,109 annotated bounding boxes in standard YOLO format, covering 4 vehicle categories: car (177,403 boxes, 82.5%), bus (3,523 boxes, 1.6%), truck (16,051 boxes, 7.5%), and van (18,132 boxes, 8.4%). The dataset is split into training set (18,655 images, 80%), validation set (2,332 images, 10%), and test set (2,332 images, 10%). All images are resized to 640x640 (native resolution 960x540). This dataset is suitable for object detection tasks, especially vehicle detection scenarios, and can be used in intelligent transportation, smart cities, and other applications. License: CC-BY-4.0.
数据集概述:Edge-AI Traffic Vehicle Detection (UA-DETRAC CCTV)
基本信息
- 数据集名称: Edge-AI Traffic Vehicle Detection (UA-DETRAC CCTV)
- 维护者:
thundarstrom - 许可证: CC-BY-4.0
- 任务类型: 目标检测(object-detection)
- 语言: 英语
- 数据规模: 10K < n < 100K
数据内容
- 图像数量: 23,319 张CCTV交通监控图像(来源于UA-DETRAC基准数据集)
- 标注框数量: 215,109 个边界框,采用YOLO标准格式
- 分辨率: 640x640(标准化处理,原始分辨率为960x540)
- 数据来源: 固定路口监控摄像头
类别分布
| 类别 | 边界框数量 | 占比 |
|---|---|---|
| 汽车(car) | 177,403 | 82.5% |
| 公交车(bus) | 3,523 | 1.6% |
| 卡车(truck) | 16,051 | 7.5% |
| 面包车(van) | 18,132 | 8.4% |
数据集划分
- 训练集: 18,655 张图像(80%)
- 验证集: 2,332 张图像(10%)
- 测试集: 2,332 张图像(10%)
使用方式
通过Python SDK下载
python from huggingface_hub import snapshot_download
local_path = snapshot_download( repo_id="thundarstrom/traffic-vehicle-detection", repo_type="dataset", local_dir="data/datasets/vehicle_detection" )
推荐训练配置
bash python scripts/train_models.py --task vehicle --model yolov8s.pt --epochs 100 --imgsz 640 --batch 32
所属项目
该数据集是 Edge-AI Real-Time Traffic Violation Detection & ANPR(边缘AI实时交通违规检测与车牌识别)项目的一部分,属于交通与车辆分析系统。




