Inspiration
The alarming rise in gun violence within educational institutions across the U.S. over the past two decades motivated us to develop an emergency response system that is both highly effective and cost-efficient.
What it does
Our system provides a comprehensive emergency response solution for gun violence in schools, leveraging only the mobile phones of students and staff, along with the existing CCTV infrastructure. It delivers accurate emergency alerts and evacuation routes through three key components.
Location tracking of mobile devices and cameras
The app prompts students to input their weekly schedules, allowing us to track which room each student is in at any given time. Additionally, the room locations of the CCTV cameras are stored in our database, ensuring all devices are precisely mapped.
Machine learning for gun violence detection
Our lightweight audio classification model uses the phone’s microphone to detect gunshots, while our object detection model analyzes CCTV footage. These systems work together to pinpoint the shooter's location based on where the mobile devices or CCTV cameras are positioned.
Real-time dashboard and push notification alerts
We use MapKit along with a custom floor plan of the building to identify rooms where potential threats are detected. Everyone in the school receives immediate alerts with precise location data. Additionally, Push notifications are sent to students and emergency responders, enhancing overall safety.
How we built it
- iOS app development using SwiftUI and MapKit
- Backend powered by SQLite, Flask and Python
- Custom audio classification model built with Apple Create ML and Core ML
- Object detection model created with Roboflow and Python
Challenges we ran into
- Integrating all platforms within a 24-hour window
- Handling real-time data rendering issues with MapKit
Accomplishments that we're proud of
- Successfully embedding our custom audio classification model into the Swift app
- Developing a tailored object detection model for analyzing CCTV footage
- Designing an accurate floor plan of the building
What we learned
- Building custom CoreML models using Apple Create ML
- Developing custom object detection models through Roboflow
- Implementing custom annotations in SwiftUI and Apple Maps
What's next for ProtectEd
- Expanding support to Android devices
- Implementing more advanced authentication for user accounts
Log in or sign up for Devpost to join the conversation.