cati-singapore-dataset
收藏资源简介:
CATI新加坡高速公路交通数据集是一个实时车辆检测数据集,数据来自新加坡陆路交通管理局(LTA)的90个交通摄像头。该数据集使用CATI(上下文感知交通智能)系统收集,该系统基于改进的YOLOv11检测器,能够实时适应环境条件。数据集包含每台摄像头的车辆检测结果,涵盖车辆数量、类别细分、方向分流和环境上下文等信息。数据集覆盖新加坡多条高速公路,包括CTE、PIE、AYE等,共约90个摄像头。数据模式包括时间戳、摄像头ID、道路代码、经纬度、天气、总车辆数、各方向车辆数、车辆类型(如汽车、摩托车、公交车等)以及检测参数(如置信度阈值、IoU阈值等)。该数据集适用于智能城市、交通流量分析、车辆检测等计算机视觉任务。数据采集间隔约为90秒,采用MIT许可证,数据源自新加坡的开放数据API。
The CATI Singapore Highway Traffic Dataset is a real-time vehicle detection dataset, with data collected from 90 traffic cameras operated by the Land Transport Authority (LTA) of Singapore. This dataset is gathered using the CATI (Context-Aware Traffic Intelligence) system, which leverages an improved YOLOv11 detector and can adapt to environmental conditions in real time. The dataset includes vehicle detection results from each camera, encompassing vehicle quantity, category breakdown, direction classification, environmental context and other relevant information. It covers multiple highways across Singapore, including CTE, PIE, AYE and others, with a total of approximately 90 cameras deployed. The data modalities consist of timestamp, camera ID, road code, latitude and longitude, weather conditions, total vehicle count, vehicle count per direction, vehicle types (e.g., cars, motorcycles, buses, etc.) and detection parameters (e.g., confidence threshold, IoU threshold, etc.). This dataset is applicable to computer vision tasks such as smart city development, traffic flow analysis and vehicle detection. The data collection interval is approximately 90 seconds, and the dataset is licensed under the MIT License, with data sourced from Singapore's open data API.
CATI新加坡高速公路交通数据集概述
数据集基本信息
- 数据集名称: CATI Singapore Expressway Traffic Dataset
- 发布者: SuhxsReddy
- 发布日期: 2026年
- 许可证: MIT
- 语言: 英语
- 数据规模: 10K<n<100K
数据集描述
该数据集包含从新加坡陆路交通管理局高速公路摄像头网络持续收集的每摄像头车辆检测结果。每条记录捕获单个摄像头的完整检测扫描,包括车辆数量、类别细分、方向分流和环境上下文。
任务与标签
- 任务类别: 目标检测
- 任务ID: 车辆检测
- 标签: traffic, singapore, expressway, smart-city, vehicle-detection, computer-vision, lta
数据覆盖范围
数据覆盖新加坡以下高速公路的约90个LTA交通摄像头:
- CTE (Central Expressway): ~16个摄像头
- PIE (Pan-Island Expressway): ~20个摄像头
- AYE (Ayer Rajah Expressway): ~8个摄像头
- ECP (East Coast Parkway): ~10个摄像头
- MCE (Marina Coastal Expressway): ~4个摄像头
- TPE (Tampines Expressway): ~8个摄像头
- BKE (Bukit Timah Expressway): ~6个摄像头
- KJE (Kranji Expressway): ~6个摄像头
- SLE (Seletar Expressway): ~6个摄像头
数据模式
| 列名 | 类型 | 描述 |
|---|---|---|
timestamp |
string | 检测时间戳 (SGT, UTC+8) |
camera_id |
int | LTA摄像头ID |
road |
string | 高速公路代码 (CTE, PIE, AYE等) |
lat |
float | 摄像头纬度 (WGS84) |
lon |
float | 摄像头经度 (WGS84) |
weather |
string | 捕获时新加坡中部天气状况 |
total_vehicles |
int | 检测到的车辆总数 |
dir_a |
int | 朝A方向行驶的车辆 (通过2帧IoU追踪) |
dir_b |
int | 朝B方向行驶的车辆 |
car |
int | 检测到的汽车 |
motorcycle |
int | 检测到的摩托车 |
bus |
int | 检测到的公交车 |
truck |
int | 检测到的卡车 |
van |
int | 检测到的厢式货车 |
lorry |
int | 检测到的货车 |
conf_threshold |
float | 使用的检测置信度阈值 |
iou_threshold |
float | 使用的NMS IoU阈值 |
imgsz |
int | 推理图像大小 (像素) |
model_version |
string | 模型检查点标识符 |
dir_a_label |
string | 人类可读的A方向标签 (例如"towards Woodlands") |
dir_b_label |
string | 人类可读的B方向标签 (例如"towards City") |
is_ramp |
bool | 如果摄像头可能在入口/出口匝道上则为True (邻近启发式) |
检测模型
检测由CATI模型生成,这是一种新颖的架构,将FiLM层注入YOLOv11s中,根据实时环境元数据调节主干网络:
- 天气状况和温度 (data.gov.sg)
- 一天中的时间 (循环编码)
- 摄像头GPS位置 (正弦位置编码)
- PM2.5空气质量
- 摄像头分辨率类别
模型权重: https://huggingface.co/SuhxsReddy/cati-singapore
数据加载方式
python from datasets import load_dataset import pandas as pd
ds = load_dataset("SuhxsReddy/cati-singapore-dataset") df = ds["train"].to_pandas()
按道路统计车辆
print(df.groupby("road")["total_vehicles"].mean())
CTE上的方向流量
cte = df[df["road"] == "CTE"] print(cte[["timestamp", "camera_id", "dir_a", "dir_b", "total_vehicles"]])
数据收集信息
- 扫描间隔: ~90秒 (完整网络扫描)
- 检测阈值: 置信度 ≥ 0.08, IoU ≤ 0.25
- 图像来源: LTA Datamall Traffic Images API
- 基础设施: HuggingFace Spaces (CPU), 持续24/7收集
数据来源与许可
- 数据来源: 新加坡开放数据API
- 数据许可: 新加坡开放数据许可证 (https://data.gov.sg/open-data-licence)
引用格式
bibtex @dataset{cati_singapore_2026, author = {SuhxsReddy}, title = {CATI Singapore Expressway Traffic Dataset}, year = {2026}, publisher = {HuggingFace}, url = {https://huggingface.co/datasets/SuhxsReddy/cati-singapore-dataset} }




