AI-Powered Video Safety


DriveTrust’s AI Dual Dash Camera is a modular build camera optimized for detection of objects from and within moving vehicles (cars, trucks, trains). Due to its modularity and powerful processing capabilities, it can detect a variety of different objects and behaviours in different scenarios, for example:

  • drowsiness and distraction in drivers
  • different types of objects in front of vehicles: people, snow on roads, blocked rails
  • distance estimation for detection of tailgating
  • usage of build-in accelerometer and GNSS for geofencing, aggressiveness, speeding and road anomalies detection

The main advantage of using the DriveTrust camera is that the processing of the video happens in real-time on the device and that no video upload is required. Therefore real-time warnings are possible in areas with little or no mobile network coverage.



The DriveTrust Score is calculated based on five different profiles, the generic, aggressiveness, drifting, tailgating and reaction profile. We use the combination of different types of data to calculate each of the profiles. The generic and aggressiveness profile can be calculated from GNSS and IMU, which is already possible with many different types of solutions. The innovation of DriveTrust is in the drifting, tailgating and reaction profile, where we combine the data from the camera and real-time object detection with the GNSS and IMU to more precisely assess the driver’s capabilities. Each profile and metric is carefully designed with the help of our car insurance partners.  More precise information on the DriveTrust Score calculation can be found here.


DriveTrust AI Camera provides a real-time track monitoring solution to help maintain and predict potential issues on railway tracks. We train our algorithms to automatically detect particular events on the tracks and to notify the corresponding railway and track maintainers companies. This could be for example:

  • dense vegetation
  • leftover equipment
  • trash
  • unidentified obstacles
  • “near-hit” incidents involving people or animals being too close to the tracks

One of our main value propositions is the ability for the customer to retrain the detection algorithms over time – we offer a simple interface where the user can reinforce the detection accuracy through “upvoting” correct and “downvoting” false detections. In the long run, this offers high detection accuracy and significantly lower costs of track problem detection and therefore its maintenance.