SentriScan

Model detecting a gun
Alert sent to mail

Inspiration

School shootings in America being a huge problem, namely the MSU shooting on Feb. 13th

What it does

It uses the custom trained YOLOv8 object detection model to identify guns and and gather data as well as the image of the gun and gunner. This data is stored in a database and retrieved when it sends the email of the threat.

How we built it

We trained a YOLOv8 model using Modal Docs API and used that model to power or object detection. To send and pull the data of the threat to and from our database, we used MongoDB API Drivers.

Challenges we ran into

We tried to train the model with a cpu and that was extremely slow. We had to learn more about how model are trained and utilize resources to train them. When we found Modal Docs as a training resource we ran into problems with making sure the model gets the right training file directories, learning the documentation of Modal Docs and dealing with idle timeouts. Also we originally wanted to create a react app with this project but through git and the complexity of the react app, we couldn't implement it into the application.

Accomplishments that we're proud of

We are proud of watching the model train in real time, as learning tons of aspects about AI object detection in just 24 hours. We are also proud of the MongoDB system we created to store the data and finding out methods to encode an image to be stored within a .json format