Second Sight

User Interface

Inspiration

We were inspired by the need for affordable, accessible security that leverages AI. By reusing old devices, we can transform everyday gadgets into smart surveillance tools, reducing waste and cost.

What it does

Second Sight repurposes old phones and laptops into AI-powered surveillance cameras. It streams video, detects motion with computer vision, and uses natural language AI to generate captions and trigger alerts for events like intrusions or package deliveries.

How we built it

Our project uses Next.js to deliver a modern, server-rendered web interface for real-time video monitoring and alert notifications. On the backend, OpenCV processes live feeds for motion detection while WebSockets provide low-latency streaming. When motion is detected, Google Gemini translates the video into natural language captions. TogetherAI is then used both to convert these captions into vector embeddings and to run our retrieval-augmented generation (RAG) model for intelligent insights. All video data—blobs, captions, and embeddings—is stored in ActiveLoop, enabling fast nearest neighbor (similarity) searches and semantic querying. This integrated tech stack transforms outdated devices into smart security solutions while reducing e-waste.

Challenges we ran into

Key challenges included achieving real-time video processing on older devices, managing varied video quality, and integrating multiple AI services seamlessly. Balancing latency with accurate AI predictions was a complex but rewarding hurdle.

Accomplishments that we're proud of

We're proud to have created a scalable, AI-powered system that gives new life to unused devices. Our solution can accurately detect and describe events using cutting-edge AI, making high-quality surveillance affordable for everyone. Our team was not too familiar with agents going into the competition so we're also proud of the way we integrated agents into our project with multiple technologies.

What we learned

We learned how to blend computer vision, natural language processing, and real-time streaming into a unified pipeline.

What's next for Second Sight

Next, we plan to enhance our AI algorithms for even better event recognition, integrate predictive analytics, and use technology like HappyRobot to call users in real-time in emergencies like a break-in or relative in distress.