|
Descripción
|
A specifically designed dataset generated using a driving path in Madrid which has been traveled 55 times using 3 different car models. The journey includes two urban areas and a connecting motorway. The overall length is 8.1 km and includes several road elements of interest (traffic lights, street crossings and roundabouts). The dataset has been generated from one driver driving the 3 different vehicles without being aware of the purpose of the experiment. He was asked to drive normally while an Android mobile device recorded the GPS traces.
Vehicles used for the data gathering.
Model and Times used:
Peugeot 206, 7
Citroen Xsara Picasso, 28
Opel Zafira, 20
|
|
Publicación relacionada
| Muñoz-Organero, M., Ruiz-Blázquez, R., Sánchez-Fernández, L. (2018). “Automatic detection of traffic lights, street crossings and urban roundabouts combining outlier detection and deep learning classification techniques based on GPS traces while driving”. Computers, Environment and Urban Systems, 68, pp. 1–8.
doi: 10.1016/j.compenvurbsys.2017.09.005 |
|
Notas
| The dataset could help reproducing the results in the companiog paper and to foster new research.
Description of the project: The dataset has been gathered to valitade the research presented in Munoz-Organero, M., Ruiz-Blaquez, R., Sánchez-Fernández, L. (2018). “Automatic detection of traffic lights, street crossings and urban roundabouts combining outlier detection and deep learning classification techniques based on GPS traces while driving”. Computers, Environment and Urban Systems, 68, pp. 1–8, https://doi.org/10.1016/j.compenvurbsys.2017.09.005
This paper combines outlier detection with deep learning based pattern recognition and classification from GPS derived data while driving. Only GPS data is used in order to minimize the deployment requirements and therefore facilitate the use of the proposed algorithm in crowdsensing scenarios. Speed and acceleration data series are computed form the GPS data. Window-segments of a constant duration around locations classified as outliers while driving are used to train a DBN (Deep Believe Network) followed by a classifier in order to detect 3 particular road elements: traffic lights, street crossings (where two roads cross each other with signs only) and roundabouts. The algorithm is applied to 16.97 h (61,076 s.) of unlabeled driving data in order to assess the detection performance.
Methodology: The GPS sensor in a mobile device is used to obtain both the vehicle's location and the estimated driving speed. The speed can be derived from the distance traveled per time unit. The distance traveled between two points 1 and 2 can be calculated form the GPS coordinates |
