🎥 Violation Detection Video App

This fullstack AI-powered web app enables users to upload and analyze videos for raised-hand violations using PoseNet. It consists of a Node.js backend and a React frontend, with MinIO for video storage and PostgreSQL for data management.

👉 You can check this video to know how it works here

---

🛠️ Tech Stack

• Backend: Node.js + Express + TensorFlow.js + MinIO
• Frontend: React + Tailwind CSS
• AI: PoseNet model for body keypoint detection
• Database: PostgreSQL
• Storage: MinIO (S3-compatible object storage)

---

📁 Project Structure

project-root/
├── frontend/
│   ├── src/
│   └── ...
├── backend/
│   ├── src/
│   └── ...
└── README.md

---

🧩 Prerequisites

• Node.js v18+
• PostgreSQL (running locally)
• MinIO (manual startup)
• FFmpeg installed globally
• TensorFlow native libraries (for Windows setup)

---

🔧 Backend Setup

1. Install dependencies

   cd backend
   npm install

2. Environment variables: Create a .env file with the following:

   MINIO_ACCESS_KEY=minioadmin
   MINIO_SECRET_KEY=minioadmin
   MINIO_ENDPOINT=127.0.0.1
   MINIO_PORT=9000
   MINIO_USE_SSL=false
   
   PG_USER=postgres
   PG_HOST=127.0.0.1
   PG_DATABASE=video_recording_db
   PG_PASSWORD=postgres
   PG_PORT=5432
   
   PORT=5000

3. Initialize the database

   npm run init-db

4. Start the backend server

   npm start

---

💻 Frontend Setup

1. Navigate to the frontend directory:

   cd frontend

2. Install dependencies:

   npm install

3. Run the development server:

   npm start

The frontend runs on http://localhost:3000.

---

📦 Starting MinIO Manually

minio server /path/to/data --console-address ":9001"

---

🧠 TensorFlow Setup for Windows

1. Download libtensorflow-cpu-windows-x86_64-2.9.1.zip.

2. Copy tensorflow.dll and tensorflow.lib to:

   node_modules/@tensorflow/tfjs-node/deps/lib

---

🎞️ FFmpeg Setup (Windows)

1. Download FFmpeg from ffmpeg.org.
2. Extract it to a folder, e.g., C:\ffmpeg.
3. Add the bin folder to your system PATH (e.g., C:\ffmpeg\bin).
4. Restart your terminal or system to apply the change.

Verify installation with:

ffmpeg -version

---

🧠 How Violation Detection Works

• Extracts the first frame from the uploaded video using FFmpeg.
• Runs PoseNet on the frame to estimate body keypoints.
• Flags a violation if:
  • leftWrist or rightWrist is detected above a vertical threshold (Y < 500).
  • Detection confidence is too low (uncertain violation).
• All violations are stored in PostgreSQL.

---

✅ Scripts

Backend

npm install       # Install dependencies
npm run init-db   # Initialize the database
npm start         # Start the backend server

Frontend

npm install       # Install dependencies
npm start         # Run the React app