Crime Catcher

Inspiration

We built Crime Catcher after realizing that although cameras are everywhere, true security is often missing. Critical moments still slip through because manual human supervision for each and every screen is not practically possible at once. We wanted to turn passive recording systems into intelligent ones that not only capture footage, but also actively understand, analyze, and respond. Crime Catcher is built to work in large public spaces so that all stakeholders, from busy malls to large hospitals, stay safer in real time.

What it does

Crime Catcher converts ordinary CCTV streams into an active, AI-powered monitoring system that detects threats as they happen. Live footage is continuously processed using OpenCV for motion detection and frame extraction, then suspicious frames are analyzed and verified by a custom model on Eyepop.AI. When an event such as violence, abnormal movement, fainting, or choking is confirmed, the system automatically captures evidence, generates a timestamped report, logs it to a dashboard, and pushes instant alerts to security personnel. Instead of reviewing hours of recordings, teams receive only actionable insights in real time.

How we built it

We built the backend in Python, using OpenCV for video capture, preprocessing, and motion-based filtering to reduce unnecessary frames. Flagged images are sent to Eyepop.AI for higher-accuracy classification. A React frontend serves as the admin dashboard, displaying live feeds, incident logs, and alerts in a clean interface. The system is modular, with separate services for detection, verification, logging, and notifications, allowing us to scale across multiple cameras and maintain performance.

Challenges we ran into

Real-time stream monitoring introduced many performance and practicality bottlenecks. Inherent limitations and latencies of EyePop.ai image analysis technology also proved to be challenging. Another challenge was minimizing false positives while maintaining sensitivity, which required tuning thresholds and refining our AI verification pipeline. Integrating backend detection with frontend updates and instant alerts also required multiple synchronization attempts.

Accomplishments that we're proud of

We are proud to work on building a complete, end-to-end pipeline that works reliably from detection to notification. The system not only identifies incidents but also automatically saves evidence and generates reports. Creating a solution that could realistically help security staff respond faster and drastically improve safety was a major milestone.

What we learned

We gained practical experience in computer vision, asynchronous processing, API integrations, and frontend-backend communication. Optimizing our custom EyePop.ai model to work for various edge cases was also critical.

What's next for Crime Catcher

Emergency Service Integration: Automatically dispatch GPS coordinates to police or EMS upon detecting life-threatening events.
Multi-Camera Tracking: Re-identify and track suspects across different camera feeds in a large facility.
Audio Analysis: Incorporate sound detection (e.g., screaming, breaking glass) to corroborate visual data.