Overview

The problem being addressed is the efficient identification of trash objects within images to support waste management and environmental initiatives. By implementing an object detection model, this solution enables automated recognition of trash items, facilitating the removal and recycling of litter, thus contributing to cleaner and more sustainable communities. The solution to the problem stated above is to deploy and train the YOLOv8 model using comprehensive trash training datasets sourced from Kaggle. The trash or plastic pollution datasets will be downloaded from Kaggle. The end goal of the project will be to create a usable app to run the model for easy implementation and detection. This optimized model aims to surpass the capabilities of TrashAI. Ultimately, the project's objective is to deliver a user-friendly application for seamless implementation and efficient detection. Additionally, in the hardware domain, this model can be seamlessly integrated with a robotic system capable of detecting and recording trash data, as well as physically removing the detected trash, addressing the issue comprehensively. The primary challenges in creating this solution is creating image annotations in bulk and achieving high accuracy trash and plastic pollution detection while maintaining efficient performance.

Project Advising

Dr. Eric Shook, Director, GeoCommons

Michelle Andrews, GIS Professional