VisionSentinel AI

VisionSentinel AI uses OpenCV 5 and agentic AI to analyze videos, detect events, gather visual evidence, and generate explainable reports with privacy protection and AWS deployment.

Comment

Inspiration

Video surveillance systems can capture hours of footage, but finding important events and collecting relevant evidence can take significant time. We wanted to explore how computer vision and AI could make video investigation more efficient, explainable, and privacy-conscious.

What It Does

VisionSentinel AI is an agentic computer vision project designed to analyze video footage, identify visual events, extract timestamped evidence frames, and support structured incident investigation.

The platform brings video analysis, evidence exploration, AI-assisted investigation, evaluation tools, and report generation into one interface. Its goal is to help users move from raw video footage to organized, reviewable findings.

How We Built It

The project uses a React and TypeScript frontend, a Python FastAPI backend, and OpenCV-based video-processing functionality. The interface provides a command-center dashboard for video analysis, evidence review, investigation, and reporting. Gemini integration is intended to provide AI-assisted analysis when the API is configured and available.

Challenges We Faced

One of the main challenges was connecting the user interface with video-processing functionality while keeping results understandable and evidence-linked. Another challenge was distinguishing computer vision observations from AI-generated interpretations and handling missing services or unsuccessful processing gracefully.

What We Learned

We explored how computer vision, AI-assisted reasoning, evidence extraction, and user-friendly interfaces can work together in an investigation workflow. We also learned the importance of testing real inputs, communicating uncertainty, and designing systems that support human review rather than blindly trusting AI outputs.

What's Next

Our next steps include validating performance across more real-world videos, improving event tracking and evaluation, strengthening privacy controls, and testing reproducible deployment on AWS.

VisionSentinel AI aims to make video investigation more organized, transparent, and practical while keeping evidence and human oversight at the center.

Built With

Share this project:

Updates

Submission history