Deploy a powerful traffic monitoring model trained on a massive 880-image dataset for comprehensive urban surveillance. This project features a pre-trained computer vision model optimized to detect 10 distinct classes, including cars, buses, emergency vehicles, and pedestrians, providing the scale needed for advanced smart city infrastructure.

Ways to Use Traffic Detection Dataset

1. Traffic Flow Analysis: The dataset can be used in machine learning models to analyze traffic flow in cities. It can identify the type of vehicles on the city roads at different times of the day, helping in planning and traffic management.

2. Vehicle Class Based Toll Collection: Toll booths can use this model to automatically classify and charge vehicles based on their type, enabling a more efficient and automated system.

3. Parking Management System: Parking lot owners can use this model to easily classify vehicles as they enter for better space management. Knowing the vehicle type can help assign it to the most suitable parking spot.

4. Traffic Rule Enforcement: The dataset can be used to create a computer vision model to automatically detect any traffic violations like wrong lane driving by different vehicle types, and notify law enforcement agencies.

5. Smart Ambulance Tracking: The system can help in identifying and tracking ambulances and other emergency vehicles, enabling traffic management systems to provide priority routing during emergencies.

Traffic-related data collected by the Boston Transportation Department, as well as other City departments and State agencies. Various types of counts: Turning Movement Counts, Automated Traffic Recordings, Pedestrian Counts, Delay Studies, and Gap Studies.

~_Turning Movement Counts (TMC)_ present the number of motor vehicles, pedestrians, and cyclists passing through the particular intersection. Specific movements and crossings are recorded for all street approaches involved with the intersection. This data is used in traffic signal retiming programs and for signal requests. Counts are typically conducted for 2-, 4-, 11-, and 12-Hr periods.

~_Automated Traffic Recordings (ATR)_ record the volume of motor vehicles traveling along a particular road, measures of travel speeds, and approximations of the class of the vehicles (motorcycle, 2-axle, large box truck, bus, etc). This type of count is conducted only along a street link/corridor, to gather data between two intersections or points of interest. This data is used in travel studies, as well as to review concerns about street use, speeding, and capacity. Counts are typically conducted for 12- & 24-Hr periods.

~_Pedestrian Counts (PED)_ record the volume of individual persons crossing a given street, whether at an existing intersection or a mid-block crossing. This data is used to review concerns about crossing safety, as well as for access analysis for points of interest. Counts are typically conducted for 2-, 4-, 11-, and 12-Hr periods.

~_Delay Studies (DEL)_ measure the delay experienced by motor vehicles due to the effects of congestion. Counts are typically conducted for a 1-Hr period at a given intersection or point of intersecting vehicular traffic.

~_Gap Studies (GAP)_ record the number of gaps which are typically present between groups of vehicles traveling through an intersection or past a point on a street. This data is used to assess opportunities for pedestrians to cross the street and for analyses on vehicular “platooning”. Counts are typically conducted for a specific 1-Hr period at a single point of crossing.

A dataset containing the traffic network information in Los Angeles city from March to Jun 2012. It is used in the traffic forecasting task in Graph Neural Networks.

The source from the paper: https://arxiv.org/abs/2206.09113

The METR-LA dataset contains 02 information: - A adj_METR-LA.pkl: is the graph that contains the physical connection of 207 loop detectors in the traffic network. This dictionary contains 03 elements: the real sensor ID, the mapped sensor to node ID, and the adjacency matrix. - A METR-LA.h5: is the time series that is collected from each sensor in the traffic network over time.

Daily utilization metrics for data.lacity.org and geohub.lacity.org. Updated monthly

~150 million nationwide parcels with land use type, building density, housing units, points of interest, and job density.

Network-level vehicle miles traveled data supporting climate and transportation performance monitoring across all US geographies.

The PeMS traffic datasets have been collected by the California Transportation (Caltrans) agency for 30-second granularity, and the raw and aggregated data are publicly available on their website (https://pems.dot.ca.gov/?dnode=Clearinghouse&type=meta&district_id=7&submit=Submit). We have gathered 5-minute aggregated vehicular traffic state (i.e traffic speed) dataset for district four and seven of California for 2022.

We have used Bing Distance Matrix API to compute a driving distance between each sensor. The API can be used to compute a driving distance between a single source or multiple sources and source or multiple destinations at once.

In addition, the weather datasets have been collected from https://www.visualcrossing.com/weather/weather-data-services and the datasets have one-hour granularity, and we have only removed some of the unnecessary columns.

You can also access an API version of this dataset.

TMS

(traffic monitoring system) daily-updated traffic counts API

Important note: due to the size of this dataset, you won't be able to open it fully in Excel. Use notepad / R / any software package which can open more than a million rows.

Data reuse caveats: as per license.

Data quality

statement: please read the accompanying user manual, explaining:

how

 this data is collected identification 

 of count stations traffic 

 monitoring technology monitoring 

 hierarchy and conventions typical 

 survey specification data 

 calculation TMS 

 operation. 

Traffic

monitoring for state highways: user manual

[PDF 465 KB]

The data is at daily granularity. However, the actual update

frequency of the data depends on the contract the site falls within. For telemetry

sites it's once a week on a Wednesday. Some regional sites are fortnightly, and

some monthly or quarterly. Some are only 4 weeks a year, with timing depending

on contractors’ programme of work.

Data quality caveats: you must use this data in

conjunction with the user manual and the following caveats.

The

 road sensors used in data collection are subject to both technical errors and 

 environmental interference.Data 

 is compiled from a variety of sources. Accuracy may vary and the data 

 should only be used as a guide.As 

 not all road sections are monitored, a direct calculation of Vehicle 

 Kilometres Travelled (VKT) for a region is not possible.Data 

 is sourced from Waka Kotahi New Zealand Transport Agency TMS data.For 

 sites that use dual loops classification is by length. Vehicles with a length of less than 5.5m are 

 classed as light vehicles. Vehicles over 11m long are classed as heavy 

 vehicles. Vehicles between 5.5 and 11m are split 50:50 into light and 

 heavy.In September 2022, the National Telemetry contract was handed to a new contractor. During the handover process, due to some missing documents and aged technology, 40 of the 96 national telemetry traffic count sites went offline. Current contractor has continued to upload data from all active sites and have gradually worked to bring most offline sites back online. Please note and account for possible gaps in data from National Telemetry Sites. 

The NZTA Vehicle

Classification Relationships diagram below shows the length classification (typically dual loops) and axle classification (typically pneumatic tube counts),

and how these map to the Monetised benefits and costs manual, table A37,

page 254.

Monetised benefits and costs manual [PDF 9 MB]

For the full TMS

classification schema see Appendix A of the traffic counting manual vehicle

classification scheme (NZTA 2011), below.

Traffic monitoring for state highways: user manual [PDF 465 KB]

State highway traffic monitoring (map)

State highway traffic monitoring sites

The main aim of this dataset is to enable detection of traffic congestion from surveillance cameras using one-stage object detectors. The dataset contains congested and uncongested traffic scenes with their respective labels. This dataset is collected from different surveillance cameras video footage. To prepare the dataset frames are extracted from video sources and resized to a dimension of 500 x 500 with .jpg image format. To Annotate, the image LabelImg software has used. The format of the label is .txt with the same name as the image. The dataset is mainly prepared for YOLO Models but it can be converted to other models format.

Network traffic datasets created by Single Flow Time Series Analysis

Datasets were created for the paper: Network Traffic Classification based on Single Flow Time Series Analysis -- Josef Koumar, Karel Hynek, Tomáš Čejka -- which was published at The 19th International Conference on Network and Service Management (CNSM) 2023. Please cite usage of our datasets as:

J. Koumar, K. Hynek and T. Čejka, "Network Traffic Classification Based on Single Flow Time Series Analysis," 2023 19th International Conference on Network and Service Management (CNSM), Niagara Falls, ON, Canada, 2023, pp. 1-7, doi: 10.23919/CNSM59352.2023.10327876.

This Zenodo repository contains 23 datasets created from 15 well-known published datasets which are cited in the table below. Each dataset contains 69 features created by Time Series Analysis of Single Flow Time Series. The detailed description of features from datasets is in the file: feature_description.pdf

In the following table is a description of each dataset file:

This table provides the traffic studies in hourly bins and some statistics. The SDOT Traffic Counts group runs studies across the city to collect traffic volumes. Most studies are done with pneumatic tubes, but some come from video systems as well. Use the field study_id to match it with other tables for more information.

Here are a few use cases for this project:

1. Autonomous vehicle navigation: Utilize the "Carla traffic dataset" to train self-driving vehicles in detecting vehicles, pedestrians, traffic signs, and traffic lights, enabling them to navigate safely and adhere to traffic regulations.

2. Traffic analysis and management: Implement the dataset to create a smart traffic management system capable of analyzing vehicular and pedestrian movement while adjusting traffic light timings for optimal flow and reduced congestion.

3. Surveillance and security: Integrate the dataset with CCTV cameras and security systems to monitor and detect unusual activities, such as pedestrians or bikers entering restricted areas, as well as violations of traffic rules.

4. Urban planning and infrastructure development: Use the data to analyze pedestrian and vehicle movement patterns, identifying areas requiring improved infrastructure, such as additional bike lanes, crosswalks, or traffic control features.

5. Augmented reality for navigation: Incorporate the "Carla traffic dataset" within AR applications to provide real-time information on traffic conditions, nearby pedestrians, bikers, and traffic signs, enhancing user's navigation and transportation experiences.

Driving behavior signals and crash records supporting road safety analysis, vulnerable road user exposure assessment, and high-risk location identification.

Overview

The Traffic System Dataset provides a collection of key metrics for analyzing and optimizing traffic flow patterns across multiple lanes and intersections. It can be used for predictive modeling, congestion analysis, and intelligent traffic control system design. This dataset includes real-time measurements such as vehicle counts, average speed, lane occupancy, and waiting time, making it ideal for researchers, data scientists, and urban mobility engineers working on smart city and transportation analytics.

Context

Efficient traffic management is one of the most crucial challenges in modern urban planning. With the growth of smart cities, AI and data-driven solutions have become essential for monitoring traffic flow, predicting congestion, and reducing waiting times. This dataset captures multiple time-based and performance-related parameters of a traffic system, providing a foundation for: • Predictive traffic control systems • Vehicle flow optimization • Intelligent transportation system (ITS) modeling • Reinforcement learning applications in traffic light scheduling

Dataset Features

Feature Name Description vehicle_count Number of vehicles passing through a specific observation point during a given time interval. average_speed Mean speed (in km/h or mph) of all vehicles detected in the observation period. lane_occupancy Percentage of lane space occupied by vehicles, indicating traffic density. flow_rate Rate of vehicle flow per unit time (e.g., vehicles per minute). time_of_day Time label or categorical feature representing different traffic periods (e.g., morning peak, afternoon, evening). waiting_time Average waiting time (in seconds) for vehicles during signal cycles or congestion periods.

respectively.

This table provides traffic studies in 15 minutes bins and some statistics. The SDOT Traffic Counts group runs studies across the city to collect traffic volumes. Most studies are done with pneumatic tubes, but some come from video systems as well. Use the field study_id to match it with other tables for more information.

Copy of https://universe.roboflow.com/roboflow-100/vehicles-q0x2v Labels reduction is applied to improve models performance and reduce class imbalance: - All busses are united in class bus - All busses are united in class truck

This dataset aims to analyze the traffic flow patterns in Dhaka, focusing on both vehicular movement and pedestrian activities. Data were gathered from four different locations: Shapla Chattar, Arambag, Bashabo, and Abul Hotel. Video recordings were taken from footover bridges, capturing traffic scenarios involving single-lane and double-lane roads, as well as the erratic movement of pedestrians. A total of 23,678 images were extracted from these recordings, which were collected during five distinct time intervals on a weekday, and subsequently annotated using the Roboflow tool.

Data in Brief Paper Link: https://www.sciencedirect.com/science/article/pii/S2352340925001301

Location of traffic counters in York. *Please note that the data published within this dataset is a live API link to CYC's GIS server. Any changes made to the master copy of the data will be immediately reflected in the resources of this dataset.The date shown in the "Last Updated" field of each GIS resource reflects when the data was first published.