Home / Current Issue / Paper 1717898
Deep Learning-Based Smart Waste Segregation Using Image Classification
Subject area: Science,Engineering and Technology · Area of research: AI and Deep Learning / Smart Waste Management
DOI: https://doi.org/10.64388/IREV9I11-1717898
Abstract
Rapid urbanization and rising levels of consumption have been a major contributor to the amount of solid waste that is produced globally and this has created a serious environmental and health risk to the population. The inefficient source segregation of waste is one of the main bottlenecks of the modern waste management systems. The conventional sorting techniques are extremely time consuming, dangerous to the worker and can easily be subject to human errors. Over the past few years, scholars have studied the concept of artificial intelligence and deep learning to automate the process of waste classification by using image recognition, and they have suggested models utilizing Convolutional Neural Networks (CNNs) and transfer learning frameworks, including ResNet and MobileNet, to enhance classification performance.Nevertheless, the available literature can be prone to small datasets, complex and computationally intensive models, or controlled laboratory settings, which impair their application in real-world smart waste management systems to a considerable degree. Most of the proposed methods are also not scalable and cannot be implemented in real-time, which makes them inappropriate when it comes to integrating them into embedded or edge computing devices that are prevalent in the smart city infrastructure.In this paper, the review of the recent deep learning solutions to the problem of waste image classification will be provided, and the main research gaps in the literature will be identified. The research will use the YOLOv8 (You Only Look Once) architecture as an automated waste segregation system based on the lightweight deep learning framework, suggested in light of the gaps identified. In contrast to two-stage detectors, the single-stage object detection pipeline of YOLO has the ability to localize and classify objects in multiple classes (e.g., plastic, paper, glass, and metal, and organic waste) at once, making it a real-time inference system, which can be used to deploy smart bins and conveyor-belt sorting devices.The suggested framework is educated and assessed with the help of the wide variety of augmented waste picture data and estimated with the help of typical measures such as mean Average Precision (mAP), precision, recall, and performance speed. The system is designed to achieve a high degree of classification accuracy with a computational efficiency that is efficient enough to be used in real time applications and, thus, facilitates efficient recycling operations, less manual sorting and effort, and leads to sustainable smart city waste management systems.
Keywords
Deep Learning, Waste Classification, Smart Waste Management, Image Classification, Convolutional Neural Networks (CNN), Transfer Learning, Smart Cities
References
[1] K. Gomathi and L. K. Narayanan, “A Deep Learning Aided Smart Waste Classification System for Smart Cities,” in Proc. IEEE Int. Conf. Computational Intelligence and Communication Networks (CICN), 2024, pp. 160–164.
[2] N. Singh, K. S. Reddy, and R. Aluvalu, “AI Driven Waste Classification for Smart Recycling,” in Proc. IEEE Int. Conf. Disruptive Technologies (ICDT), 2025, pp. 1590–1594.
[3] P. P. Sanjana, Y. Krishna, A. M. P., L. Jayakrishna, and A. K. Sivaramakrishnan, “AI-Based Waste Classification Using ResNet50 and PyTorch: A Practical Approach for Real-World Deployment,” in Proc. IEEE Int. Conf. Women in Innovation, Technology & Entrepreneurship (ICWITE), 2025.
[4] V. Balaji, P. Perichetty, M. Giri, and M. R. Sekhar, “Analysis of Garbage Classification with a Convolutional Neural Network,” in Proc. IEEE Int. Conf. Communication, Computing and Energy Efficient Technologies (I3CEET), 2024, pp. 1064–1069.
[5] J. P. Bhati, C. Garg, H. K. Saran, and A. Singh, “Automated Trash Detection and Classification Using CNN: A Deep Learning Approach,” in Proc. IEEE Int. Conf. Networks and Cryptology (NETCRYPT), 2025, pp. 1030–1034.
[6] M. S. Hossen, A. J. Shoukhin, N. H. Athin, and A. Chakma, “Automated Waste Management: A Deep Learning and Cloud-Based Food & Plastic Waste Separator Bin,” in Proc. IEEE Int. Conf. Sustainable Technologies for Industry 5.0 (STI), 2024.
[7] R. Kashaf, E. P. Alegre, T. Prova, and S. Aggarwal, “Automated Waste Management Using a Customized Vision-Based Transformer Model,” in Proc. IEEE World AI IoT Congress (AIIoT), 2024.
[8] A. Kovitvadhi, P. Sapsoonthorn, S. Thainimit, and P. Rattanadechkamjohn, “YOLO-Based Waste Classification and Detection for Smart Recycling,” in Proc. IEEE Int. Smart Cities Conf. (ISC2), 2024.
[9] A. Narendran, S. Sundaram, A. H. Inuguri, and M. Yashaswini, “Smart Waste Management System: Automating Waste Sorting with Image Classification and IoT Integration,” in Proc. IEEE Int. Conf. Compute, Control, Network & Photonics (ICCCNP), 2025.
[10] C. Sirawattananon, N. Muangnak, and W. Pukdee, “Designing of IoT-Based Smart Waste Sorting System with Image-Based Deep Learning Applications,” in Proc. IEEE ECTI Conf., 2021, pp. 383–387.
[11] A. M. Nasir and R. Roslan, “Real-Time Recycle Waste Classification Using Deep Learning: A CNN Model for Automated Waste Management,” in Proc. IEEE Int. Conf. Computing (ICOCO), 2024, pp. 440–445.
[12] D. A. Janeera, P. R., S. V., and M. S. Sheela, “Visual Servoing and Deep Capsule Network Learning for Contactless Smart Waste Segregation,” in Proc. IEEE Int. Conf. Smart Electronics and Communication (ICOSEC), 2021, pp. 1301–1305.
How to cite this paper
@article{1717898,
author = {Hatim Murtaza Ramiwala, Dr. Balaurugan S.},
title = {Deep Learning-Based Smart Waste Segregation Using Image Classification},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {11},
pages = {2378-2385},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1717898.pdf},
abstract = {Rapid urbanization and rising levels of consumption have been a major contributor to the amount of solid waste that is produced globally and this has created a serious environmental and health risk to the population. The inefficient source segregation of waste is one of the main bottlenecks of the modern waste management systems. The conventional sorting techniques are extremely time consuming, dangerous to the worker and can easily be subject to human errors. Over the past few years, scholars have studied the concept of artificial intelligence and deep learning to automate the process of waste classification by using image recognition, and they have suggested models utilizing Convolutional Neural Networks (CNNs) and transfer learning frameworks, including ResNet and MobileNet, to enhance classification performance.Nevertheless, the available literature can be prone to small datasets, complex and computationally intensive models, or controlled laboratory settings, which impair their application in real-world smart waste management systems to a considerable degree. Most of the proposed methods are also not scalable and cannot be implemented in real-time, which makes them inappropriate when it comes to integrating them into embedded or edge computing devices that are prevalent in the smart city infrastructure.In this paper, the review of the recent deep learning solutions to the problem of waste image classification will be provided, and the main research gaps in the literature will be identified. The research will use the YOLOv8 (You Only Look Once) architecture as an automated waste segregation system based on the lightweight deep learning framework, suggested in light of the gaps identified. In contrast to two-stage detectors, the single-stage object detection pipeline of YOLO has the ability to localize and classify objects in multiple classes (e.g., plastic, paper, glass, and metal, and organic waste) at once, making it a real-time inference system, which can be used to deploy smart bins and conveyor-belt sorting devices.The suggested framework is educated and assessed with the help of the wide variety of augmented waste picture data and estimated with the help of typical measures such as mean Average Precision (mAP), precision, recall, and performance speed. The system is designed to achieve a high degree of classification accuracy with a computational efficiency that is efficient enough to be used in real time applications and, thus, facilitates efficient recycling operations, less manual sorting and effort, and leads to sustainable smart city waste management systems.},
keywords = {Deep Learning, Waste Classification, Smart Waste Management, Image Classification, Convolutional Neural Networks (CNN), Transfer Learning, Smart Cities},
month = {May},
doi = {https://doi.org/10.64388/IREV9I11-1717898}
}