The trio of Luis Oliver, Victor Izquierdo and Alejandro Gutiérrez won the Jetson Project of the Month for their Drowsiness, Blindspot and Emotion Monitor (DBSE). This project, powered by NVIDIA Jetson Nano, is an in-car assistance system that alerts the driver if they’re drowsy or distracted and notifies them about objects in their blindspot.
The functionality of the system is divided among Drowsiness detection, Emotion Detection and Driving Monitor (using Yolov3) modules. The drowsiness and emotion modules use OpenCV’s Haar Cascades approach for face detection. Once the driver’s face is detected, the modules use convolutional neural networks built in PyTorch and running on Jetson Nano to detect the state of the driver’s eye (open/close) and the driver’s emotion. In the Driving Monitor module, objects in a driver’s blind spot are identified using the Yolov3 algorithm (trained on the 80-class COCO dataset).
All the modules send visual or audio alerts based on the detected information using the MQTT protocol to either the mini-OLED display or the speaker. For instance, the mini-OLED display shows the type of object in the blindspot and LEDs on the left and right of the display indicate the relative position of the object. Similarly, the speaker sounds an alarm when the driver is distracted.
Per NHTSA, 400,000 people have been injured in crashes involving distracted drivers in 2018 and 4,111 fatalities occurred in motor vehicle crashes involving drowsy driving between 2013 and 2017. While there’s no alternative to being alert and focused at the wheel, in-car assistance systems such as DBSE aim to provide an additional level of safety to the driver, passengers and others on the road. Luis, Victor and Alejandro plan to improve the project and believe that this has the potential to be a viable commercial solution. We can’t wait to see how this project evolves.
For developers and users to build their own versions of this system, the team has open-sourced the bill of materials and the code here.
