Internship Project: Automated System for Waste Identification and Separation

Introduction

This project was developed in collaboration with the Instituto Superior de Engenharia do Porto (ISEP), as part of the PESTI (Internship Project) course. The main objective of the project is to create an automated system using computer vision and artificial intelligence (AI) for identifying and sorting recyclable waste, such as plastic, glass, paper, and metal. Through the implementation of this system, we aim not only to improve the efficiency of recycling but also to provide a solid framework for future projects related to waste management and environmental sustainability.

Project Objectives

The primary goal of this project is to provide educational material that enables users to replicate, train, create, and understand the development of AI models for waste classification. This material will be valuable for schools and communities, offering a strong foundation for the development of future projects. The developed system will be accompanied by step-by-step guides, practical examples, and educational resources, serving as a framework for the creation and adaptation of AI solutions for similar projects in the future.

YOLOv8 Training & Inference Scripts for Image Classification, Object Detection, and Segmentation

This repository serves as a comprehensive guide for training and deploying YOLOv8 models across three major computer vision tasks:

• Image Classification: Assigns a single class label to an entire image, identifying the dominant object or scene (e.g., "plastic", "glass", "metal","cardboard","paper","styrofoam", etc).

• Object Detection: Locates and classifies multiple objects within an image using bounding boxes (e.g., detecting and labeling several recyclable materials in one photo).

• Instance Segmentation: Extends object detection by generating precise pixel masks for each detected object, enabling more accurate shape and contour recognition.

The scripts allow you to train custom models on your own datasets and generate both visual outputs (e.g., annotated images with boxes and masks) and text-based results for use in real-time systems, analytics, or integration into larger applications.

Developed by

Afonso Guilherme Vieira da Silva Oliveira
Student Number: 1221160