AI-Powered Vision Assistance Prototype

For our FSE 100 project at Arizona State University my team was tasked to create a solution for a group of people. After hearing the interviews that we were provided we realized that people with vision impairments have a hard time with head-high objects and wanted to make a solution to help warn them before they hit an obstacle

Our solution involved a combination of a webcam and ultrasonic sensor. The ultrasonic sensor warned of an imminent danger with a loud buzzer sound while the webcam was connected to an OpenAI API which detected if there were any obstacles farther out which then warned the user with a vibration. The whole thing had a button to turn it off should the user be sitting on public transit where there would be many close obstacles. We used async Python functions to allow the ultrasonic sensor to continue to work while the webcam was waiting for a response

The compute was done on a Raspberry Pi 4B and the main hardware was stored in a fanny pack. The ultrasonic sensor and webcam were housed on top of a baseball cap with the wires carefully managed.

The end result impressed the professor enough for him to keep our design around for an example for next year