Project Documentation & Demo

Project Report (PDF/Word): Report

Demo Videos: Video Demo

Project Overview

This project is an AI-powered Personal Protective Equipment (PPE) violation detection system designed to help enterprises improve workplace safety monitoring through real-time computer vision and automated alerts.

The system uses deep learning (YOLO-based object detection) to identify whether workers comply with mandatory safety requirements such as:

• Hard hats

• Safety vests

• Face masks

When a violation is detected, the system can automatically send alert emails with evidence images/videos to supervisors or safety managers.

This solution is suitable for:

• Construction sites

• Manufacturing plants

• Warehouses

• Industrial zones

• Smart factory environments

Project Objectives

The main goals of this project are:

• Reduce workplace accidents caused by PPE violations

• Automate safety supervision using AI instead of manual monitoring

• Provide real-time detection from images, videos, and live webcam streams

• Store inspection history for auditing and reporting

• Notify responsible personnel immediately via email

Key Features

AI-Based PPE Detection

• Detects:

  • Hardhat / No-Hardhat

  • Safety Vest / No-Safety Vest

  • Mask / No-Mask

• Built on YOLO deep learning architecture

• Custom-trained model (ppe.pt)

Multi-Input Support

Users can perform detection using:

• Uploaded images

• Uploaded videos

• Real-time webcam stream

User Authentication System

• Secure login & sign-up system

• Password hashing

• Session-based authentication

• Individual user history tracking

Detection History Management

• Stores:

  • Uploaded files

  • Processed outputs

  • Detection statistics

  • Violation count

  • Timestamp

• Powered by MongoDB database

Email Alert System

• Automatic email notifications when violations occur

• Includes:

  • Violation summary

  • Detection time

  • Evidence reference

Web-Based Platform

• Flask backend

• HTML/CSS frontend

• Works entirely through browser

• No software installation required for end users

System Architecture

User
│
│ Upload Image / Video / Webcam
▼
Web Interface (Flask)
│
│ Authentication & Validation
▼
AI Detection Engine (YOLOv8)
│
│ Bounding box + classification
▼
Result Processor
│
├── Save result to MongoDB
├── Store media in server
└── Trigger email alert

Technology Stack

Backend

• Scipy

• Python 3.10+

• Flask

• PyTorch

• Matplotlib

• OpenCV

• Ultralytics

• YOLOv8

• Google Colab

Frontend

• HTML5

• CSS3

• Bootstrap

• Javascript

Database

• MongoDB (Local / MongoDB Atlas)

AI & Computer Vision

• YOLO object detection

• Custom PPE dataset

• Trained .pt model

Example Detection Output

• Bounding boxes with confidence score

• Violation highlighted in red

• PPE-compliant objects highlighted in green

• Summary statistics:

  • Number of workers

  • PPE compliance count

  • Violation count

 

Potential Business Applications

• Smart construction monitoring

• Factory safety automation

• AI-powered CCTV systems

• Industrial IoT integration

• Safety compliance analytics dashboard

Future Enhancements

• OSHA rule-based violation scoring

• SMS / Slack / Microsoft Teams alerts

• Multi-camera CCTV integration

• Real-time dashboard analytics

• Cloud deployment (AWS / Azure / GCP)

• Edge AI deployment (Jetson Nano / Orin)

💻 Developer

Khoi Bao
Computer Science Student

Core Skills

• Computer Vision

• Deep Learning

• Python Backend

• Flask & REST API

• MongoDB

• Web AI Deployment

📬 Contact

For collaboration, research, or enterprise integration:

📧 Email: khoibao655@gmail.com
🌐 GitHub: https://github.com/KhoiBao1

Summary

This project demonstrates the ability to:

• Build a real-world AI safety system

• Integrate deep learning with full-stack web development

• Deploy AI models into production-ready web applications

• Design solutions aligned with U.S. industrial safety standards