SafePath Shield

Inspiration

Traditional school safety systems react too slowly and too broadly. When an emergency occurs, sprinklers activate across the entire building, lockdowns affect all rooms, and vital information is often delayed or confusing. We wanted to design an AI-powered safety platform that acts intelligently, locally, and automatically — protecting lives while minimizing unnecessary damage or panic.

What It Does

SafePath Shield is a web-based platform that combines computer vision, dynamic mapping, and real-time automation to make buildings safer, smarter, and more efficient.

Dynamic Floor Mapping

Users can upload or draw a floor plan of their building.
They can place nodes in each room and label them (e.g., “hallway,” “classroom,” “bathroom”).
A machine learning model detects boundaries such as walls and doors, automatically dividing the layout into dynamic rooms.

Real-Time Threat Detection (YOLO-based)

Each webcam feed connects to a YOLO-based model trained on large datasets for fire, gun, and knife detection. The system can manage multiple webcams simultaneously, analyzing each in real time. When a threat is detected, the affected room node is immediately highlighted on the map.
Smart Fire Response

If fire is detected, SafePath Shield activates sprinklers only in that room, reducing water waste and preventing unnecessary damage.

Smart Security Response

If a weapon is detected, nearby rooms automatically lock, and warning notifications are sent to building occupants.
The system also generates dynamic, color-coded exit routes using arrows to guide people to safety.

Interactive 2D Demo

The demonstration view shows a real-time 2D layout of the building.
When a threat occurs, affected areas turn red, exit paths update dynamically, and automated responses are visualized in real time.

How We Built It

Frontend: React.js with Canvas rendering for floor plan interactivity
Backend: Flask or FastAPI server for camera management and model inference
Machine Learning:
- YOLOv8 models trained on open-source fire and weapon detection datasets
- Custom-labeled data for improved indoor accuracy
- Computer Vision and Mapping: OpenCV for wall detection and boundary creation from floor plan images
- Communication: WebSocket-based update system for real-time alerts and visualization

Impact

Cost-effective: Activates sprinklers only in affected rooms, reducing water and repair costs.
Rapid Response: AI enables faster detection and action than manual reporting.
Scalable: Works with any number of webcams and any building layout.
Adaptable: Can be deployed in schools, hospitals, corporate offices, or industrial facilities.

Challenges We Ran Into