Metagros

• 

• 

  Multi-camera surveillance with real-time YOLOv8 object detection, loitering alerts and crowd, and AI-powered behavior analysis

• 

• 

  Secure authentication portal.

• 

  Natural language video search powered by Twelve Labs. Upload footage and describe incidents to find timestamped evidence clips.

Inspiration

We wanted to make security monitoring more effective and less overwhelming by helping teams focus on what actually matters.

What it does

• Reduces the time and effort needed to review surveillance footage.
• Performs real-time object detection on live video using YOLO.
• Allows natural language search over recorded footage.

How we built it

• OpenCV
• YOLO (Ultralytics)
• Twelve Labs API
• QT Framework

Challenges we ran into

• Cursor and Antigravity crashing :D
• Managing real time performance across video streams

Accomplishments that we're proud of

• Successfully combining live detection with semantic video search
• Building a working desktop prototype within the hackathon timeframe

What we learned

• How to combine traditional computer vision with modern video understanding models
• How to design a system that balances performance, accuracy, and usability
• How to develop desktop applications

What's next for Metagros

• Multi camera support
• Testing with real security camera feeds
• Secure remote access and monitoring

Built With

• opencv
• pyside6
• python
• sqlite
• twelvelabs
• yolo