Utilizing our school's CCTV cameras, we developed a weapon detection system, which would notify administration and local authorities to be able to adequately handle a situation.
Once our MVP was built out, we pitched our idea. Our demo helped us gain approval and traction amongst our schools principal, districts tech department, and superintendent. Working alongside these individuals helped expose us to opportunities to deploy our project.
For more info regarding our tech stack, please see:
Our first test site was our district's warehouse which alleviated the privacy concern associated with working with security cameras—on a student campus.
For training purposes, we began by detecting common objects like phones, water bottles, and hats. Once we were more confident in the model’s training, we then transitioned to training for prop weapons.
Once we proved these models were successful we were granted permission to use a CCTV camera at the back of our school.
The project is continuing to progress as we continue to train and fine-tune the ML-model. Additionally, we’re also looking to improve the speed of our system and quality of the footage by testing various codecs (vp80, h264, etc).
In terms of expansion, we’re working with our superintendent to began planning out a larger scale beta-testing. Schools in our district have hundreds of cameras located throughout the campus—we’re working with the district to determine which cameras would be most effective for the system.
In addition, we are exploring more functionality than just weapon detection, such as, detecting animals, and tracking how many people are not wearing masks.
Lastly, we are making an effort to work with our city's police department to see how we can better cater our project to help them respond to threats.