Team 08 - Baikely

Every year, 45,000 cyclists are involved in car accidents. The only common solutions to these dangers are reflective vests and flashing lights, where bikers have to blindly trust cars to see them first. But bikers want to safely see their surroundings, and these solutions only help bikers be seen. That is why we have built Baikely to put your safety back into your control.

Baikely is a bicycle safety system designed to provide you with greater sense of awareness of your surroundings. Our product uses a rear-mounted vehicle detection system to search for and identify vehicles behind your bike. Then, after complex software analysis, our product delivers the locations of vehicles, bicycles, and other obstacles on a handlebar display. Once you’re aware of the vehicles behind you, you’ll be safer and have more confidence while biking. When you turn Baikely on, it begins to look for vehicles and obstacles creeping up behind you. When Baikely detects something, you'll hear a voice alert such as “car left” and if a vehicle gets too close, you'll hear a loud beep. Baikely presents everything on an intuitive GUI display, providing you a multisensory experience. Lastly, Baikely is easily installable, designed to fit on any bike.

Our vehicle detection system is centered around a camera. Data from the camera is processed with a computer vision model to detect cars and calculate their position relative to the biker. We derived this model from OpenCV and optimized it to run on the low-power Raspberry Pi with TensorFlow Lite. Our system also holds 3 ultrasonic sensors to measure the exact distance to nearby vehicles or objects. This adds an additional layer of precision, particularly in areas just outside of the camera’s field of view as this component has 270-degree coverage. Over on the display, our product shows the results of its computer vision analysis and sensors. We used Pygame to render a simple view of your surroundings to make it easy to see where vehicles and bicycles are located behind you. It shows both the locations of the vehicles and the ultrasonic sensor readings. Together, this system provides a refreshingly simple user interface. Installation takes just three minutes. Our 3D printed design fits easily on bikes of any shape, thanks to our flexible back camera mechanism. Once installed, Baikely will offer two modes - urban and suburban. Choose your mode, and Baikely will adjust its detection range in response to how crowded your environment is.

The main hardware components of Baikely are the raspberry pi, ultrasonic sensors, and the raspberry pi camera. The 3.5 inch display is compact, allowing it to be mounted on any handlebar. Furthermore, with our custom 3D-printed casing, all sensors fit snugly underneath the bike seat. In terms of software, we used Onshape and Fusion 360 for 3D modeling, TensorFlow Lite and OpenCv for vehicle detection, PyGame for our GUI, and eSpeak for text to speech. In total, unit costs for our prototype’s hardware total to 142 dollars. However, we believe that with mass production costs it can be reduced to just 60 dollars.