Inspiration
Distracted driving is one of the leading causes of road accidents each year in Singapore and the rest of the world
15% of all motor vehicle crashes in 2021 were due to distracted driving
On average, nine people die every day from distracted driving
Hence we came up with an idea to prevent this issue
What it does
Leveraged computer vision to detect the concentration of the driver. In case the driver is looking away or has eyes closed for more than a few seconds, an alarm will be sounded with blinking LED within few seconds to wake the driver up or to remind the driver to pay attention and stay alert.
The vision detection system uses a scoring system and on a high enough score(~>4) the screen turns red and sends a signal input to our alarm system built using arduino. The arduino takes in the signal and sounds an alarm and causes the LED to blink.
How we built it
Used a dataset of 2000 images of eye-movements.
Implemented a light-weight CNN model, we trained the model using keras and Tensorflow.
Used CV2(Open CV) for real time driver face tracking and eye movement score prediction.
Built arudino alarm to buzz an alarm whenever the score is beyond a threshold to alert the driver
Challenges we ran into
Making accurate predictions to avoid false positives
Integrating CV with arduino to ensure proper functioning of alarm and make sure no unneccessary noise created
Accomplishments that we're proud of
The model has an accuracy of 96.88% in predicting the driver's distractedness. Can quickly alert once by tracking the eye features efficiently.
What we learned
Fine tuning our CNN models to provide better accuracy. Integrating different systems to provide meaningful and impactful solutions.
What's next for Distracted-Driver-Detection
Can be extended to mobile environments and edge devices in various long journey vehicle systems due to its computational cost effectiveness.
Log in or sign up for Devpost to join the conversation.