WatchTower

Inspiration

With rising concerns around safety and security, we wanted to build a system that could spot criminal activity in real time. Many surveillance systems only record incidents, but few can actively analyze and alert nearby people in real time.

What it does

WatchTower uses Gemini AI alongside OpenCV to process video feeds. The system detects suspicious or criminal acts, flags them, and provides a way to monitor places more intelligently. Instead of passively recording, WatchTower actively interprets what it sees

How we built it

1. Integrated OpenCV for video capture and frame analysis
2. Used Gemini AI to classify and detect potential criminal actions
3. Built a lightweight Python GUI for visualization and monitoring
4. Designed modular code so new detection features can be added easily

Challenges we ran into

1. Integrating Gemini’s API smoothly with OpenCV video streams
2. Handling performance bottlenecks when analyzing high-resolution video in real time
3. Balancing detection accuracy with processing speed

Accomplishments that we're proud of

1. Successfully connected Gemini AI with OpenCV in a working prototype
2. Built an end-to-end system that detects, processes, and displays events in real time
3. Delivered a functional project that addresses real-world problems

What we learned

1. The importance of data preprocessing and real-time performance optimizations
2. Collaboration and task-splitting under tight deadlines

What's next for WatchTower

1. Deployment on surveillance cameras
2. Expanded detection library (move beyond criminal acts to also recognize accidents, fires, and other emergencies).
3. Ethical safeguard

Built With

• gemini
• opencv
• python