Excited to share a project I helped build during Hack Hazards '25: an AI-powered web application that detects potential stampede risks from live or uploaded videos in real time. 🚨 The Problem: Crowd stampedes claim hundreds of lives each year. Whether it’s concerts, festivals, or transit hubs, detecting and acting on crowd surges early is critical. 💡 Our Solution: Stampede Predictor uses computer vision and streaming data to detect and visualize high-risk crowd zones, enabling real-time alerts for organizers and public safety teams. 🔧 Key Features: ✅ Real-time camera + video analysis ✅ YOLOv8-powered person detection ✅ Crowd density heatmap grid (Normal → Critical) ✅ Live SSE risk updates on UI ✅ Audio alerts for CRITICAL risk ✅ Fluvio message streaming pipeline ✅ Downloadable processed video with overlays 🛠️ Stack & Tools: • YOLOv8 Nano (Ultralytics) • Flask + OpenCV (Python) • HTML/CSS/JavaScript (Frontend) • Fluvio (event streaming) • SSE (live UI updates) • Threading + Queues (live status sharing) 🏁 My Contributions: I worked on: • Flask backend architecture • YOLO-based detection + risk grid logic • Fluvio integration for real-time streaming • Server-Sent Events (SSE) for live UI updates • Audio alert trigger system • Deployment readiness and modular structure 🏆 Top 15 — Fluvio Track, Hack Hazards '25 This project placed among the Top 15 in the Fluvio Track (from 17,000+ hackers across 25+ countries). Huge thanks to my teammates and The NAMESPACE Community for the opportunity! 📂 GitHub: github.com/gabsgj/Stampede-Predictor 🎥 Demo: youtu.be/KKmF_QUh2yI 📽️ Devfolio: devfolio.co/projects/stampede-predictor 👨💻 Team Arete: Gabriel James · Nayana Shaji · Jany Sabarinath · Vrindha P #HackHazards25 #Fluvio #YOLOv8 #ComputerVision #CrowdSafety #OpenCV #StreamingAI #HackathonProject #AIForGood #Python #RealtimeAI #GECTThrissur #TeamArete