Supervisor

Dr. Muhammad Iqbal, Ken Healy

Programme

BSc (Hons) in Computing in IT

Subject

Computer Science

Abstract

Waste contamination is a major problem all over the world. The Environmental Protection Agency reports that over two thirds of waste found in general household and commercial bins could have been placed in the recycling or organic waste bins instead in Ireland, with food waste and plastics being the most common misplaced items. This project presents a deep learning solution to classify nine categories of waste from camera images in real time with the goal of helping users sort waste correctly at the source. Following the CRISP DM framework, two convolutional neural network architectures were trained and compared, MobilenetV2 and Resnet 18 both using transfer learning with frozen feature extractors. MobilenetV2 was selected as the better baseline with 80.25 percent test accuracy and a macro F1 of 0.808, outperforming Resnet 18 across nearly every class while also being smaller, faster and more suitable for edge deployment. After testing the baseline in real conditions with a webcam the model was improved through a two phase experiment first stratified splitting with class weights and then fine tuning of the last three feature blocks. The methodology changes alone contributed roughly 1.3 percentage points while fine tuning contributed an additional 7.5 taking the final model to 89.76 percent test accuracy and a macro F1 of 0.899. The selected model was deployed using OpenCV onto my Laptop and then to Raspberry Pi 5 to confirm it could run on edge hardware. Both deployments work without retraining and the live demo correctly classifies common household waste items.

Date of Award

2026

Full Publication Date

2026

Access Rights

open access

Document Type

Undergraduate Project

Resource Type

bachelor thesis

Publisher

https://doi.org/10.63227/652.299.149

Download

Share

COinS