SecureSight

• 

  SecureSight Dashboard

Inspiration

I was motivated by growing concerns surrounding accountability and transparency in law enforcement interactions. Recent controversies involving altered or lost body-cam footage inspired me to combine AI-based detection with blockchain to ensure crucial evidence remains tamper-proof and verifiable. My goal was to create a decentralized, reliable record that fosters public trust.

How I Built It

I developed the frontend using React, creating a real-time webcam interface powered by TensorFlow.js. The AI models used include COCO-SSD for detecting people and weapons, and MoveNet for recognizing specific actions like "hands up". The backend was built using Solidity, with an Ethereum smart contract deployed via Hardhat. Using Ethers.js, I seamlessly connected frontend detections directly to blockchain storage, allowing each AI-identified event to be immutably logged.

Challenges I Faced

Integrating real-time AI detection with blockchain posed latency challenges, requiring creative solutions like event batching. Fine-tuning the in-browser AI models to accurately and quickly identify events took considerable trial and error. Managing multiple concurrent detections (objects and poses) was complex yet highly rewarding.

What I Learned

This project significantly enhanced my understanding of blockchain integration with real-time AI, teaching me valuable lessons in optimizing performance, accuracy, and user experience within decentralized applications.

Built With

• cocossd
• ethereum
• javascript
• movenet
• react.js
• solidity
• tensorflow.js