NORMA eResearch @NCI Library

A Comparative Analysis for Trash image classification using Deep Learning

Mysore Dayananda, Prithvi (2022) A Comparative Analysis for Trash image classification using Deep Learning. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (2MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


Garbage classification is the most essential tool in the given era. Over 3.40 billion tonnes of waste is predicted to be produced by the year 2050. At this rate the toll garbage production has on the environment, health, marine, and wildlife are enormous. In this fast-moving technological era, we civilians must follow the three R’s of waste management, recycle, reuse and reduce. Climate change for one is a great example that we take precautions and try to build a better future. In recycling waste, the most crucial part is of segregating the right material. Since a mixture of materials is of no use. Currently, there are sensors used to divide materials but the machines fail to categorize all of them which then leads to manual handpicking of trash. To solve this we have implemented multiple convolutional neural networks such as VGG16, ResNet50, and a custom model MLH-CNN. A comparative study has been done on all of the models mentioned. Poor or random initialization of parameters may lead to longer training or vanishing gradient problems. Hence in this research, many experiments have been implemented and the results have been compared. The main objective of this research is to compare the performance of the light-weighted custom MLH-CNN model with VGG16 and ResNet. Since the MLH-CNN model was trained on TrashNet data with only 2575 images only the simple structure worked well, the dataset used in this research is 15728 images and by multiple experiments, we see that the model’s accuracy has been depreciated and ResNet50 outperforms Custom MLH-CNN and VGG16 model with an accuracy of 82.20% with an overall precision and recall of 85.30% and 80.36% respectively.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
T Technology > TD Environmental technology. Sanitary engineering
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 23 Feb 2023 17:44
Last Modified: 02 Mar 2023 08:31

Actions (login required)

View Item View Item