Canada, University of Toronto/University of Waterloo
Kathrine Elizabeth Von Friedl
Sleep deprivation is known to be an important risk factor for car crashes around the world. Drowsy driving not only puts the drivers at risk but also risks the lives of everyone else on the road. If a driver was to be sleepy and distracted and remain unresponsive behind the wheel, they would most likely cause an accident on the road. In an effort to reduce the number of sleep-related crashes and save lives, our team is working on the Blink project. Blink uses Microsoft's custom vision AI to train a neural network model using labeled photos of open, closed and tired eyes. Using a webcam, companies such as Uber, Lyft and various taxi services may have the ability to monitor their drivers (specifically whether they become sleepy/unresponsive). Captured video shots are then given to our trained model and a label is sent back to our backend. When Blink detects that the user's eyes have been closed or sleepy for 3s, the system issues an alert using Twilio. Immediately the latitude and longitude of the driver are forwarded to an emergency contact of the driver, telling them the situation of the driver and the location of the car. This can be updated when the service is being used by taxi companies in a way that the company can manage not to suggest new passengers to the driver and ask them to stop driving. This can build trust in customers of such services, and they are assured to not be at risk of any accidents due to the driver’s tiredness. We are hoping that with the expansion of such a system, risk of motor vehicle accidents caused by drowsy driving is reduced and car crashes may be reduced.
SafeTrip was shaped as a group during the UofTHacks IV which was a hackathon hosted by the University of Toronto in late January 2019. We were thrilled to learn about Microsoft Azure and how we can advance our project using it. Particularly, we were fond of the custom vision AI and were excited to see how flexible and easy to use the software was. In the past few years, the recent improvements of artificial intelligence and computer vision have been a great asset to identifying expressions from someone’s face and our team decided to use this as a starting point of creating a system which would reduce the risk of accidents due to drowsiness by identifying sleepy drivers using computer vision and neural networks. Our project was selected as the top 5 groups of the entire hackathon among around 120 other groups and we also won the Microsoft Azure Champ prize. This was the first step which led us to consider taking the project to next levels and to submit it to Imagine Cup. We are now pursuing our idea mainly because we want to take initiative in safe driving. Impaired driving is regretfully becoming a common event. Through drinking or simply being phone addicted, people are jeopardizing themselves and others on the road. A massive 93% of car crashes are avoidable. With a simple but well implemented and constructed software, we hope to implement a means of reducing car crash fatalities. Our team consists of 3 members, Elaheh Jalalpour is a Computer Science Masters alumni from the University of Waterloo (graduated Sep 2018). Her degree focus is in Computer Networks and her thesis articulates around applied AI for intrusion detection. She has a solid background in python back-end development and she is currently working as a Machine Learning Engineer in Aviva Canada. Partoo Vafaeikia is a 4thyear computer science student at the University of Toronto and will graduate in April 2019. Her degree focus is in Artificial intelligence is also doing her minor in Statistics. She has taken many courses in the field of AI and have done various projects in machine learning and computer vision. These skills will be very helpful in this project since one of the main technical expertise we need is in image processing and neural networks. In addition, she has a solid experience in full-stack development and have been part of the product management team at Densify as an intern. Another teammate is Kathrine von Friedl. Kathrine has currently finished her first term pursuing her studies of Mechatronics Engineering at the University of Waterloo. Her co-op program will finish in April 2023. She is very passionate about machine learning and robotics evident in her founding of a local robotics club at her high school. Kathrine has taken up many side projects including SafeTrip as well as projects dealing with drone development, industrial robotic programming (in a model factory), virtual and augmented reality development and development on various hardware. Kathrine is currently on a co-op work term as a virtual reality developer and research assistant, spending her free time at hackathons and aiding the University of Waterloo’s robotics team. Other than practicing C++ as taught in her program, Kathrine is also self teaching C# (for the Unity game engine), ROS and Python for her hacks.