Inspiration
My project was inspired by the increasing amount of accidents happening because of visually impaired individuals not understanding exactly what is in their surroundings. I felt that there can be a very simple, yet effective solution to this. Therefore, I started developing Percepta.
What it does
Percepta uses a Raspberry Pi Zero 2W to run a local object detecting AI model through a camera in combination with a Time of Flight sensor. Using python, I am able to detect exactly what the camera is detecting, how far that object is, and then send that through a Text to Speech engine to a Bluetooth audio device like a speaker or wireless earbuds.
How we built it
I built this using my previous knowledge of working with sensors, and the raspberry pi. I used Claude to code the python code. I used my 3d printer to 3d print the casing, and Ki Cad to create the PCB and solder it together.
Challenges we ran into
Some challenges I ran into were actually getting the Bluetooth to work on the Raspberry Pi. Raspberry Pi Bluetooth is notoriously finicky, but I pushed through and worked it out. Another challenge I ran into was my printer breaking midway through the print on the last day of submissions, so I had to improvise to make the shell a sandwich type and use some standoffs to secure it.
Accomplishments that we're proud of
I am proud of keeping the cost down of this product by using minimal cost components. The most expensive component was the Raspberry Pi at only around 20 dollars. The PCB is also low cost and so are the sensors. I am also proud of the accuracy of the device and how well it works.
What we learned
The biggest thing I learned from this was solving problems in a time sensitive manner. I learned to quickly resolve issues and strategize while under a time limit. I also learned about using different types of sensors and using a local AI model on a raspberry pi to detect objects.
What's next for Percepta
This device was nothing more than a prototype. I think that if this concept is expanded upon, it can provide drastic assistance to visually impaired individuals. If we can improve the compute power of the computer, and increase speeds of the AI detection, or make the TTS voice sound nicer, it can really prove to be a successful and greatly helpful product.
Built With
- 3dprinted
- ai
- batterypowered
- pcb
- raspberrypi
- timeofflight
Log in or sign up for Devpost to join the conversation.