The UAH-DriveSet is a public collection of data captured by our driving monitoring app DriveSafe by various testers in different environments. This dataset tries to facilitate progress in the field of driving analysis by providing a large amount of variables that were captured and processed by all the sensors and capabilities of a smartphone during independent driving tests. The application was run on 6 different drivers and vehicles, performing 3 different behaviors (normal, drowsy and aggressive) on two types of roads (motorway and secondary road), resulting in more than 500 minutes of naturalistic driving with its associated raw data and additional semantic information, together with the video recordings of the trips. You can download the dataset on the Download Section.
You can find more information in the paper presented in ITSC2016. If you found this dataset useful please cite us in your work:
E. Romera, L.M. Bergasa and R. Arroyo, "Need Data for Driving Behavior Analysis? Presenting the Public UAH-DriveSet", IEEE International Conference on Intelligent Transportation Systems (ITSC), pp. 387-392, Rio de Janeiro (Brazil), November 2016. [pdf]
DriveSafe is an app that monitors, scores and alerts your driving. By simply placing it in your windshield, DriveSafe processes the phone accelerometers, GPS and rear-camera to produce a complete driver evaluation and behaviour profile that can be checked either on real-time during the trip or reviewed with detail when the route has ended.
This app evaluates the phone accelerometers to detect brusqueness in your vehicle's motion, processes the position in lane and the lane changes, the maximum allowed speed and number of lanes of the road(provided by OpenStreetMap), and detects the ahead vehicles and the distance/time kept to them. This allows DriveSafe to score each trip according to 7 manoeuvres: accelerations, brakings, turnings, lane-weaving, lane-drifting, overspeeding and car-following. It also rates each trip within 3 behaviour models: normal, drowsy and aggressive.
An augmented reality interface allows you to enhance the trip experience and improve your driving by obtaining feedback of your behaviour. You can also disable augmented reality by switching to an interface that shows you any of the scores and the behaviour ratios on real-time. All the information is also available to the user after the trip has ended, including map views with the info, location and risk of the different performed manoeuvres, together with automatically recorded videos of the dangerous ones.
DriveSafe also produces a series of alerts during driving (sounds and interface signs) which otherwise would be only available in premium vehicles: sudden inertial events (accelerations, brakings turnings), lane changes (normal or irregular), prolonged zig-zag behaviour, overspeeding, and close car-following (tailgating).
Our app works with any vehicle and in adverse weather conditions like severe raining, or at night (vehicle detection performance might be reduced with lack of light). Additionally, our app is privacy-friendly and it only uses the rear camera (aiming the road) to analyse the route, without taking any image from the front camera that aims the interior of the vehicle.
If you found DriveSafe App useful in your research, please cite us in your works:
L. M. Bergasa, D. Almería, J. Almazán, J. J. Yebes and R. Arroyo,"DriveSafe: an App for Alerting Inattentive Drivers and Scoring Driving Behaviors", in IEEE Intelligent Vehicles Symposium (IV), pp. 240-245, Dearborn, Michigan (United States), June 2014. BEST POSTER AWARD. [pdf][poster]
E. Romera, L. M. Bergasa and R. Arroyo, "A Real-time Multi-scale Vehicle Detection and Tracking Approach for Smartphones", in IEEE Intelligent Transportation Systems Conference (ITSC), pp. 1298-1303, Las Palmas, Canary Islands (Spain), September 2015. [pdf]
The dataset was collected by 6 drivers with different ages and vehicles, including a fully electric vehicle. Thre behaviors (normal, drowsy and aggressive) were performed in two different routes, one is 25km (round trip) in a motorway type of road with normally 3 lanes on each direction and 120km/h of maximum allowed speed, and the other is around 16km in a secondary road of normally one lane on each direction and around 90km/h of maximum allowed speed.
A reader tool has been made available with the dataset, as there are several variables and files for every route and syncing them with the recorded video may suppose difficulties. This tool allows to select each of the routes in order to simultaneously reproduce the associated video and plot a selection of variables synced in real-time within an user interface. This tool can be used to find patterns in the driving behaviors by reviewing all the variables available in the dataset together with the videos that show what did actually happen during the tests. For example, the user may analyze how is a car-following maneuver in an aggressive behavior by reviewing the real-time plot of the distance that the drivers keep with respect to the ahead vehicle.
NEW: Download the tool from the Github repository
You can download the full dataset after accepting our License agreement: