IMPACT BY THE NUMBER
precision in driving risk identification
recall in driving risk identification
A large Japanese insurance firm wanted to create an insurance plan based on the driving behavior of elderly people. Since older citizens accounted for the majority of accidents in the country, the firm required a method to identify driving behavior in the elderly to help them create the appropriate insurance plans.
To assist our client with this, Addo designed an AI Engine to monitor the driving behavior—and the associated health risks that were attached with it—of the elderly in real-time. Using infrared cameras, a number of indicators were analyzed to assess the driver’s health status, emotional mindset, and general physical ability when hard braking, turning abruptly or decelerating rapidly, to reveal patterns of risky driving. Moreover, data from the on-board diagnostics (OBD) sensors was evaluated to understand if the driver’s ability to handle the car was compromised.
We developed a scalable solution with preprocessing and prediction pipelines using a combination of data; telematics from the OBD, data from car cameras, traffic and weather information, etc. The model was trained to identify and predict when elderly drivers would begin to drive in a risky manner—analyzing indicators like Age Estimation, Mood Detection, Hand Gesture Identification, Gaze Detection, Pose Estimation—so that they could be advised to stop and park.
A team of highly qualified Machine Learning Engineers, Data Scientists and Cloud Specialists at Addo was engaged for this project.
AI Techniques: Convolutional Neural Networks, Ensemble and Boosting methods, SVM-L1 Regression, RFE-CV, Extra Trees Classifier
- Potential correlation between volatile heart rate and risky driving identified
Looking for a similar project?