VisiTrack - Facial Recognition Visitor System

VisiTrack is an AI-powered visitor tracking and security dashboard designed to manage entry logs via facial recognition.

Originally developed at IIT Kanpur (Public Policy and Opinion Cell), this application demonstrates the client-facing interface of a system capable of real-time face detection, demographic analysis, and automated logging.

---

🚀 Features

• 📊 Interactive Dashboard: Real-time visualization of visitor traffic, category breakdowns (Students, Faculty, Staff), and system health status.
• 📷 Live Face Scanner: Integrates with the device webcam to capture visitor images.
• 🤖 AI Analysis (Gemini Powered):
  • Uses Google Gemini 2.5 Flash to simulate the backend recognition engine.
  • Estimates demographics (Age, Gender, Emotion).
  • Simulates identity matching with confidence scores.
• 📝 Visitor Logs: Searchable and filterable activity log (Check-in / Denied).
• 🗄️ Database Management: Interface for browsing registered profiles (simulated 5k+ records).

---

📸 Screenshots

Dashboard View

Overview panel showing traffic analytics and system status.

---

Live Face Scanner

Webcam-based facial capture and AI result preview.

---

Visitor Logs

Entry history with filters and status markers.

---

🛠 Tech Stack

• Frontend: React 19, TypeScript
• Styling: Tailwind CSS, Lucide React
• Charts: Recharts
• AI Integration: Google GenAI SDK (@google/genai)
• Build Tool: Vite

---

ℹ️ Project Context & Architecture

Original Backend (Full System)

The real system was backed by a Python pipeline:

• Dataset Building: Web scraping IIT Kanpur directory data.
• Recognition Engine:
  • FaceNet
  • dlib
  • HOG
  • SVM
  • KNN
• Achieved 90%+ accuracy on internal data.
• Backend: Flask-based REST APIs for face encoding and verification.

---

Current Repository (Web Demo)

This repository contains only the Client-Facing Interface.

To avoid GPU dependency and deployment complexity:

• Face inference is simulated via Google Gemini API
• Gemini mimics face detection and demographic analysis
• Match confidence is AI-generated to resemble real behavior

---

📦 Usage

1. Set your Gemini API key:

   export API_KEY=your_api_key_here

2. Install dependencies:

   ```bash
   npm install
   ```
3. Run locally:

   ```bash
   npm run dev
   ```
4. Allow **camera permission** for the scanner module.

---

## 🔐 Privacy & Security

* Webcam access is used **only** for analysis.
* No image is permanently stored in the demo version.
* AI responses are session-based only.

---

## 🧑‍💻 Developer

**Adiba Khan**
*Built in May 2025*
IIT Kanpur