DriverSVT: Smartphone-Measured Vehicle Telemetry Data for Driver State Identification

The data was collected from 633 different drivers using smartphones with embedded accelerometer and gyroscope sensors while vehicle driving. The data contains more than 17.56 million samples that describe the driving behavior. The mobile devices were used to collect the accelerometer, the gyroscope, and the magnetometer data, as well as recording videos of the drivers. These videos then have been analysed by a set of deep NN-based models to calculate the driver head pose, the safety belt state, the mouth openness ratio, and to detect the dangerous state (e.g. eating, smoking, using phone, distraction,etc.) using our developed algorithms and neural netowrk models and transfered to our cloud. Additional features like the heart rate were measured using smart watch Xiaomi Mi Band 5.