🏗️🔒 ConstrucSafe AI: AI-Powered Safety Detection for Construction Sites

---

🚀 Table of Contents

• Introduction
• Features
• Technologies Used
• Installation
• Usage
• Model Training Details
• Dataset Details
• Contributing
• License
• Contact

---

📝 Introduction

Welcome to ConstrucSafe AI! 🏗️🔒

ConstrucSafe AI is a state-of-the-art application designed to enhance safety on construction sites through real-time object detection. Leveraging the power of YOLOv8, this tool detects and highlights safety equipment and potential hazards in images and videos, ensuring a safer working environment.

---

✨ Features

• Image Detection: Upload images to identify safety gear and hazards.
• Video Detection: Process and analyze videos to monitor safety compliance in real-time.
• Customizable Thresholds: Adjust confidence levels to fine-tune detection accuracy.
• Downloadable Results: Save processed images and videos with annotations for record-keeping.
• User-Friendly Interface: Intuitive design built with Streamlit for seamless user experience.

---

🛠️ Technologies Used

• Python 3.10
• Streamlit 🖥️: For building the interactive web application.
• Ultralytics YOLOv8 🤖: Advanced object detection model.
• OpenCV 📷: Image and video processing.
• NumPy 📊: Numerical computations.
• Pillow 🖼️: Image handling.
• FFmpeg 🎞️: Video encoding and processing.

---

📥 Installation

Follow these steps to set up ConstrucSafe AI on your local machine:

1. Clone the Repository

git clone https://github.com/alphatechlogics/ConstrucSafe.git
cd ConstrucSafe

2. Create a Virtual Environment

It's recommended to use a virtual environment to manage dependencies.

python3 -m venv env
source env/bin/activate

3. Install Dependencies

pip install -r requirements.txt

Note: Ensure that FFmpeg is installed on your system.

Install FFmpeg on Ubuntu

sudo apt update
sudo apt install ffmpeg

Verify FFmpeg Installation

ffmpeg -version

You should see version details if FFmpeg is installed correctly.

🎬 Usage

Run the Streamlit application using the following command:

streamlit run app.py

Once the app is running, navigate to http://localhost:8501 in your browser to access the ConstrucSafe AI interface.

Application Workflow

1. Image Detection:

   • Navigate to the Image Detection tab.
   • Upload an image (.jpg, .jpeg, .png).
   • View the original and annotated images with detected safety equipment and hazards.
   • Download the annotated image for records.

2. Video Detection:

   • Navigate to the Video Detection tab.
   • Upload a video (.mp4, .mov, .avi) within the 200MB limit.
   • The app processes the video, highlighting detected objects frame-by-frame.
   • Once processing is complete, view the annotated video directly in the app.
   • Download the processed video for further analysis.

🧠 Model Training Details

The object detection model was trained using YOLOv8 on a custom dataset tailored for construction site safety. Below are the training details:

Training Environment

• Platform: Google Colab
• Framework: Ultralytics YOLOv8
• Model Architecture: yolov8l.pt (YOLOv8 Large)

Training Command

from ultralytics import YOLO

!yolo task=detect mode=train model=yolov8l.pt data='/content/drive/MyDrive/Construction Site Safety.v30-raw-images_latestversion.yolov8/data.yaml' epochs=10

Training Output Highlights

• Total Epochs: 10
• Batch Size: 16
• Image Size: 640x640
• Optimizer: AdamW
• Loss Functions:
  • Box Loss: 7.5
  • Class Loss: 0.5
  • Distribution Focal Loss (DFL): 1.5
• Final Metrics:
  • mAP50: 0.516
  • mAP50-95: 0.333

Model Summary

• Layers: 365
• Parameters: 43,649,115
• GFLOPs: 165.5

📚 Dataset Details

The model was trained on the Construction Site Safety - v30 raw-images_latestversion dataset, curated to enhance safety monitoring on construction sites.

Dataset Source

Construction Site Safety Dataset on Roboflow

Dataset Overview

• Total Images: 717
• Annotation Format: YOLOv8
• Classes: 25 (e.g., Excavator, Gloves, Hardhat, Ladder, Mask, NO-Hardhat, NO-Mask, etc.)
• Pre-processing Applied:
  • Auto-orientation of pixel data with EXIF-orientation stripping.
• Augmentation Techniques: None applied. Accessing the Dataset The dataset was exported via Roboflow on April 19, 2023.