Computer Vision Industrial Egg Fertility Sorting System

Project Overview

This project is a YOLOv8n-based Computer Vision system for egg fertility detection, designed for high-speed, single-column conveyor belt operations in industrial settings. The system provides accurate and efficient sorting of eggs into fertile and infertile categories, optimized for speed and cost-effectiveness in a highly scalable way.

Key Features

• Pre-trained YOLOv8n model for egg fertility detection
• Optimized for high-speed conveyor belt operations
• Comprehensive testing with both individual images and batch processing
• Correction mechanism for handling multiple detections
• Designs for AWS architecture for potential future deployment
• Standarized design for implementation to per conveyor belt for high scalability

Dataset: https://www.kaggle.com/datasets/mwahyuadin/dataset-of-fertile-and-infertile-chicken-eggs

Project Structure

1. Data processing and Training: Scripts for data preparation and model training
2. YOLOv8n_egg_fertility_model: Trained YOLOv8n model for egg fertility detection
3. Testing scripts: Scripts for testing the model, including individual image and batch processing
4. AWS Implementation Design: Proposed standarized AWS architecture for highly scalable deployment

Documentation

• MODEL.md: https://github.com/DimitriVavoulisPortfolio/aws-computer-vision-industrial-egg-fertility-sorting-system/blob/main/MODEL.md
• PROCESS.md: https://github.com/DimitriVavoulisPortfolio/aws-computer-vision-industrial-egg-fertility-sorting-system/blob/main/PROCESS.md
• AWS-PLAN-DESIGN-AND-COST-REPORT.md: https://github.com/DimitriVavoulisPortfolio/aws-computer-vision-industrial-egg-fertility-sorting-system/blob/main/AWS-PLAN-DESIGN-AND-COST-REPORT.md

Model Performance

• Testing-dataset.png: https://github.com/DimitriVavoulisPortfolio/aws-computer-vision-industrial-egg-fertility-sorting-system/blob/main/Testing-dataset.PNG
• Accuracy: 1.00
• Precision: 1.00
• Recall: 1.00
• F1-score: 1.00

Speed Performance:

• Preprocess: 3.0ms per image
• Inference: 10.0ms per image
• Postprocess: 2.0ms per image
• Total processing time: 15.0ms per image

Quick Start Guide

1. Clone the repository:

   git clone https://github.com/DimitriVavoulisPortfolio/aws-computer-vision-industrial-egg-fertility-sorting-system.git
   cd aws-computer-vision-industrial-egg-fertility-sorting-system
   

2. Install dependencies:

   pip install ultralytics torch numpy opencv-python
   

3. To test the model with individual images:

   python yolov8n-egg-detection-single-image-test v1.2.py
   

4. To test the model with batch processing:

   python yolov8n-egg-detection-test-script.py
   

DISCLAIMER: The paths for the model and test images need to be specified by the user.

Future Work

• Implement AWS deployment
• Create API for real-time egg fertility prediction
• Optimize model performance for even faster processing
• Develop a user interface for system monitoring and control

License

This project is licensed under the Apache-2.0 license - see the LICENSE file for details.

Contact

For any questions or feedback, please open an issue in this repository or contact Dimitri Vavoulis.