Inspiration
Road accidents happen every day, and delays in response time can mean the difference between life and death. We wanted to create an AI-powered solution that detects crashes in real time and alerts responders immediately to minimize delays and improve safety.
What it does
We built a proof of concept that analyzes pre-recorded video footage using AI to detect crashes. When a crash is detected, it records the incident in a database and sends instant notifications to monitoring personnel. The goal is to expand this system for real-time use with CCTV cameras, enabling continuous crash detection and response in live environments.
How we built it
For the backend, we used FastAPI with Python. We utilized MongoDB to store important information and communicated with our Next.js frontend using WebSockets and REST API calls. Our AI model, trained to detect crashes from video footage, processes real-time data and triggers notifications when an incident is detected.
Challenges we ran into
Fine-tuning the AI model to minimize false positives and false negatives. Ensuring real-time processing and notifications for immediate response. Managing database updates efficiently.
Accomplishments that we're proud of
Successfully integrating AI-powered crash detection with real-time notifications. Building a scalable and efficient system for processing and storing crash data. Creating a seamless user experience that enables quick decision-making. Designing a user-friendly UI that is easy to navigate and understand.
What we learned
How to refine and train AI models effectively. The importance of handpicking and preparing quality datasets. Optimizing backend performance for real-time processing. Learned to work with Auth0 for secure authentication and user management.
What's next for SafeTrack
Expanding AI capabilities to detect crash severity and potential injuries. Integrating SafeTrack with emergency services for automatic dispatch. Enhancing system performance to support high-volume traffic surveillance. Deploying SafeTrack on public CCTV networks for real-time crash detection and response.
Log in or sign up for Devpost to join the conversation.