Object Detection for Small, Dense, and Overlapped Objects in Industrial Recycling

will be available soon...

This repository contains the code and dataset for detecting small, dense, and overlapping objects in industrial recycling processes. The focus is on enhancing object detection accuracy in challenging industrial settings using deep learning-based

Features

• Dataset: 10,000+ images and 120,000+ object instances (7 object classes).
• Comparison of Methods: Evaluation of deep learning models such as YOLO (v1-v11) and Faster R-CNN.
• Challenges Addressed:
  • Small object detection
  • Dense and overlapping object identification
  • Real vs synthetic image performance
• Data Augmentation: Techniques like flipping and mosaic augmentation are employed to improve model robustness.
• Performance Metrics: Precision, Recall, mAP, and F1-score used for model evaluation.

Installation

1. Clone this repository:

   git clone https://github.com/o-messai/SDOOD.git
   

You can copy and paste this code directly into your README.md file on GitHub. Let me know if you need any further modifications!