Secure Sight

Empowering Safety Through Every Frame

Comment 5

Inspiration

In this hackathon, we wanted to develop a solution that improves public and personal safety using AI. We are aware that increasing security concerns in public areas cause a lot of disturbance and we saw an opportunity to use real-time AI detection to help prevent threats. We all want to feel safe, and by using our AI system that quickly detects danger and alerts the proper people provides a strategic approach to security. Our SecureSight platform leverages AWS's powerful cloud infrastructure and AI services, particularly AWS Bedrock LLM, to create an integrated safety ecosystem. By combining real-time weapon detection, pose estimation for fall detection, smoke and fire monitoring, and incident reporting with instant Twilio alerts, we've built a comprehensive security solution that serves law enforcement, military defense, fire services, and first responders. The scalability and reliability of AWS EC2 and S3 services enabled us to build a robust platform capable of handling critical security operations in real-time. This project demonstrates how AWS's advanced AI capabilities can be harnessed to create practical, life-saving applications that contribute to our vision of a "Threat and Theft Free World."

What it does

We created Secure Sight, an AI platform that analyzes live video to identify potential threats and safety violations. The system detects dangerous objects and sends instant alerts to assigned contacts. That is, it can call for help even before a person realizes the danger or cannot do so themselves. We also implemented a fall detection system that identifies the pose of a person in a real-time video streams and automatically detect if the person is in the process of falling or has already fallen. The system will generate alerts along with supporting evidence to notify security admins. In addition to already mentioned threat detection, Secure Sight also includes fire and smoke detection, an AI-driven chatbot that gives information about the safety queries regarding specific area, and an incident form to report an suspicious activity or safety concerns in your area.

How we built it

In this project we used an AI-driven object detection to process live video and identify potential threats with moderate accuracy. The backend was developed using Flask, while frontend was built with React. We leveraged multiple AWS services to create a robust and scalable solution: AWS S3 buckets for secure storage of surveillance footage and incident evidence, AWS EC2 instances to host our application and handle computational workloads, AWS Lambda functions to trigger real-time alerts based on detection events, and Amazon Bedrock to power our intelligent safety chatbot with advanced language capabilities. Twilio was integrated to send real-time notifications allowing relevant authorities to respond immediately to detected threats, falls, or fire hazards. For frontend dependency management, NPM ensured efficient package handling. Our architecture demonstrates the power of AWS's comprehensive cloud ecosystem, showing how services like EC2, S3, Lambda, and Bedrock can be orchestrated to create an end-to-end safety solution that processes video streams, analyzes content for threats, stores critical data, and communicates alerts—all within a seamless, responsive system.

Challenges we ran into

Throughout the development of the SecureSight platform, we encountered several significant challenges. The integration of WhatsApp messaging via Twilio presented particular difficulties, including content type compatibility issues with uploaded media and WhatsApp Sandbox restrictions requiring recipient opt-in. We also faced API configuration complexities when connecting our React frontend with both Node.js and Python backend services, especially when transitioning from local development to production deployment on EC2. Cross-origin resource sharing (CORS) and proper handling of multimedia content in form submissions required careful implementation. Additionally, ensuring real-time geolocation functionality while maintaining user privacy presented both technical and ethical considerations. Despite these obstacles, we successfully implemented core features including the Travel Safety Assistant chatbot and the Incident Reporting system, creating a comprehensive security platform that effectively addresses travel safety concerns.

Accomplishments that we're proud of

We are proud that we have successfully built a real-time AI-powered security system that works efficiently and provides instant alerts. The AI model performs effectively in identifying not only threats but also safety risks. The user interface is intuitive and ready to use, so users will be able to navigate through the system without any problems. We think it was a great accomplishment to develop and build a fully functional security system in a short period of time.

What we learned

In this project we learned a lot about optimizing ML models for real time use and effectively managing cloud based storage. By combining our ML models (weapons detection, smoke and fire, and fall detection) we learned how to combine everything in one model without loosing performance and accuracy.

What's next for Secure Sight

We see great potential when it comes to expanding our project to large scale environments such as schools, office buildings, and industrial sites. Additionally, this technology could be applied to Medical Imaging Analysis like advanced cell nuclei detection and segmentation or machine learning-based disease classification. Another option is to use it as an AI Wildlife Camera to identify and track species in real-time. To count populations and analyze movement patterns. Thus, this project has the potential to be adapted to various solutions across multiple industries.

Built With

Share this project:

Updates