ScanAI

Main Interface
General Object Detection with Yolov5
Trained model detecting multiple guns with confidence

Inspiration

We were all very intrigued by machine learning and AI, so we decided to incorporate it into our project. We wanted to create something involving the webcam as well, so we tied it altogether with ScanAI.

What it does

ScanAI aims to detect guns in schools and public areas to help alert authorities quickly in the event of a public shooting.

How we built it

ScanAI is built entirely out of python. Computer vision python libraries were used including OpenCV, facial_recognition, yolov5 and tkinter.

Challenges we ran into

When training models, we ran into issues of a lack of ram and a lack of training data. We also were challenged by the problem of tackling multiple faces at once.

Accomplishments that we're proud of

ScanAI is able to take imported files and detect multiple faces at once and apply facial recognition to all of them. ScanAI is highly accurate and has many features including Barack Obama facial recognition, object detection, live webcam viewing and scanning, and file uploading functionalities.

What we learned

We all learned a lot about machine learning and its capabilities. Using these modules expanded our knowledge on AI and all its possible uses.

What's next for ScanAI

Our next steps would be to improve our interface to improve user friendliness and platform compatibility. We would also want to incorporate our program with Raspberry Pi to increase its usage flexibility. Lastly, we would also want to work on improving the accuracy of the detection system by feeding it more images and feedback.