EUVIMEDEUVIMED entdecken
Crossref

ARTIFICIAL INTELLIGENCE DRIVEN WASTE SEGREGATION AND RECYCLING FRAMEWORK FOR SUSTAINABLE URBAN INFRASTRUCTURE

Kamrul Hasan Rumon, Al Amin Tanmay

Journal of Integrated Sciences · 12.06.2026 · 6 (3)

DOI: 10.67295/jis.v6i3.365 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Integrated Sciences, online erschienen am 12.06.2026. Verfasst von Kamrul Hasan Rumon und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Waste segregation is a critical component of building sustainable urban infrastructure, as improper sorting of biodegradable and non-biodegradable waste can contaminate recycling streams, reduce efficiency, and increase the dependency on landfills. In this study, we proposed EWS-Net Efficient Waste Segregation Network) as a solution to this issue. The designed framework is a deep learning-based auto-classification system for waste management that enhances recycling efficiency with the power of artificial intelligence. EWS-Net is based on EfficientNetB3, which is end-to-end trainable with dynamic learning rate adjustment for ensuring robust feature extraction and convergent stability. Two benchmark datasets from Kaggle, namely the Waste Segregation Image Dataset and the Garbage Classification Dataset (for cross-domain generalization), were utilized, and a well-ordered preprocessing chain, including class balancing, augmentation, and hierarchical structuring, was employed to prevent overfitting during training. Experimental results indicate that EWS-Net outperforms current state-of-the-art models, such as ResNet50, DenseNet121, MobileNetV2, and VGG16, with an accuracy of 97.32%, a precision of 0.9756, a recall of 0.9732, and an F1-score of 0.9739. Cross-dataset validation also indicates that the model can generalize with an accuracy of 94.1% on the Garbage Classification Dataset, which is boosted to 96.0% upon few-shot fine-tuning. This research proposes EWS-Net, a robust deep learning architecture that enables the accurate and automatic segregation of waste, addressing the critical limitations of previous manual and CNN-based approaches. With the reduction of human oversight errors, enhancement of recycling processes, and the ability to adapt to various waste datasets, the architecture itself directly supports innovative waste management systems, making the vision of smart and sustainable cities and circular economy initiatives a reality.

Zitieren

APA

Rumon, K. H., & Tanmay, A. A. (2026). ARTIFICIAL INTELLIGENCE DRIVEN WASTE SEGREGATION AND RECYCLING FRAMEWORK FOR SUSTAINABLE URBAN INFRASTRUCTURE. Journal of Integrated Sciences, 6(3). https://doi.org/10.67295/jis.v6i3.365

Vancouver

Rumon KH, Tanmay AA. ARTIFICIAL INTELLIGENCE DRIVEN WASTE SEGREGATION AND RECYCLING FRAMEWORK FOR SUSTAINABLE URBAN INFRASTRUCTURE. Journal of Integrated Sciences. 2026;6(3). doi:10.67295/jis.v6i3.365

BibTeX
@article{Rumon2026,
  author = {Rumon, Kamrul Hasan and Tanmay, Al Amin},
  title = {ARTIFICIAL INTELLIGENCE DRIVEN WASTE SEGREGATION AND RECYCLING FRAMEWORK FOR SUSTAINABLE URBAN INFRASTRUCTURE},
  journal = {Journal of Integrated Sciences},
  year = {2026},
  volume = {6},
  number = {3},
  doi = {10.67295/jis.v6i3.365}
}
Kostenlos starten →