> PassBy Analytics_

🔴 Try the Live Demo | 📺 Watch the Presentation

🚀 The Pitch

Have you ever run a booth at a conference or hackathon and wondered: "How many people actually stopped and talked to us versus just walking by?"

Enter PassBy Analytics.

PassBy is a lightweight, zero-setup computer vision application. You open a website on your laptop, point your webcam at the crowd, and our Edge AI instantly categorizes the crowd into Passersby and Engaged Leads based on mouth movement. If they are talking and asking questions, you'll know!

✨ Features That Wow

• 🗣️ Lip-Reading Engagement: We don't just track bodies. We use Google's FaceLandmarker 468-point 3D face mesh to track the exact vertical distance between the upper and lower lips. If someone's mouth is moving, we instantly flag them as an active, engaged lead!
• ⚡ Blazing Fast Edge AI: Runs two simultaneous ML models (EfficientDet-Lite0 for bodies + FaceLandmarker for lips) directly in your browser using WebAssembly. No cloud APIs. No latency.
• 🔒 Absolute Privacy: Because the AI runs entirely in the browser's memory, no images or video feeds are ever sent to a server. What happens at the booth stays at the booth.
• 🎯 Zero Hardware: Forget expensive LiDAR sensors or infrared cameras. If you have a laptop with a webcam, you have a smart analytics dashboard.

🛠️ How We Built It Under The Hood

The secret sauce is in how we combine tracking and facial landmarks:

1. Body Tracking: We draw a bounding box around everyone in frame.
2. Facial Correlation: We detect faces and calculate an algorithmic mouthOpenness score (distance between lips divided by face height, ensuring accuracy from 1 foot or 10 feet away).
3. State Management: We map the Face ID back to the Body ID using bounding-box intersection. If the mouthOpenness score rapidly fluctuates above 0.05 for 5 consecutive frames, our categorizer promotes that person from PASSING to ENGAGED.

💻 Run It Locally (In 60 Seconds)

You want to see the magic yourself? It's plug-and-play.

# 1. Clone the repo
git clone https://github.com/Metroxe/websummit-hackathon.git
cd websummit-hackathon

# 2. Install Dependencies
npm install

# 3. Start the Vite Dev Server
npm run dev

Open http://localhost:5173/websummit-hackathon/ in your browser. Allow webcam access. Stand in front of the camera, and start talking!

---

Built with ☕ and ❤️ at the Web Summit Hackathon.