Automatic Lightweight CNN Waste Identification for Green Campus Program Support
DOI:
https://doi.org/10.64803/ikosstemi.v1.31Keywords:
Lightweight CNN, MobileNetV2, Waste Classification, Green Campus, Artificial IntelligenceAbstract
Effective waste management is a key indicator of success for Green Campus Programs and a major factor in the UI GreenMetric World University Rankings. However, many Indonesian universities still face difficulties due to low accuracy in manual waste sorting, particularly for inorganic and hazardous (B3) waste. This study develops a lightweight, high-accuracy automatic waste classification system using the MobileNetV2 Convolutional Neural Network (CNN) architecture. The model was trained via transfer learning on the Kaggle “Trash Classification Dataset” (Sathish, 2023) containing 19,762 images from 10 waste classes: Metal, Glass, Biological, Paper, Plastic, Cardboard, Battery, Shoes, Clothes, and Trash.
To align with operational needs of the Green Campus Program, these 10 classes were mapped into three functional categories: Organic, Recyclable, and Inorganic/B3. Experimental results show that MobileNetV2 achieved 93.2% accuracy with efficient inference time (~4.8 ms per image) on a CPU. The prototype, built using Python Streamlit, outputs both predicted waste type and confidence percentage, making it practical for real-time campus waste sorting. The proposed model provides an intelligent, energy-efficient, and transparent solution to support sustainable waste management, addressing key operational challenges in inorganic (WS4) and hazardous (B3) waste (WS5) handling
References
Al-Khatib, I. A. (2023). Deep learning-based approach for accurate and real-time classification of waste. Sustainable Cities and Society, 95(104618), 1–12.
Adiputra, R. (2023). Waste management analysis at Universitas Negeri Jakarta (UNJ) using UI GreenMetric indicators. Indonesian Journal of Science and Humanities, 4(1), 1–15.
Habibie, M. L. (2022). The application of machine learning using Google Earth Engine for remote sensing analysis. Jurnal Teknoinfo, 16(2), 233. https://doi.org/10.33365/jti.v16i2.1872
Islam, M. M., Hasan, S. M. M., Hossain, M. R., Uddin, M. P., & Mamun, M. A. (2025). Towards sustainable solutions: Effective waste classification framework via enhanced deep convolutional neural networks. PLOS ONE. https://doi.org/10.1371/journal.pone.0324294
Nugroho, A. (2024). Implementation of MobileNetV2 for automated waste segregation in smart cities utilizing trustworthy AI framework. Journal of Energy-Efficient Systems, 10(1), 45–56.
Ouchra, H., Belangour, A., & Erraissi, A. (2024). Supervised machine learning algorithms for land cover classification in Casablanca, Morocco. Ingénierie Des Systèmes D Information, 29(1), 377–387. https://doi.org/10.18280/isi.290137
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). MobileNetV2: Inverted residuals and linear bottlenecks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 4510–4520.
Sathish, S. (2023). Trash Classification Dataset. Kaggle.
Sanyal, S., et al. (2024). Comparative optimization of EfficientNetB3, MobileNetV2, and ResNet50 for waste classification. Journal of Applied Informatics and Technology, 5(2), 110–125.
Susilo, B. A. (2024). CNN Lightweight MobileNetV2: Comparison of parameter efficiency against VGG16 for image classification. International Journal of Computer Vision and Robotics, 19(3), 201–215.
Yong, L., Ma, L., Sun, D., & Du, L. (2023). Application of MobileNetV2 to waste classification. PLOS ONE, 18(3), 1–16. https://doi.org/10.1371/journal.pone.0282336
Zhao, Q., Chen, S., & Li, Q. (2023). Research on lightweight convolutional neural network in garbage classification. International Journal of Computer Engineering and Technology, 14(4), 1–11.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Nouval Khairi, Yuda Apriyansyah, Haikal Habibi Siregar, Muhammad Farhan Aditya (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
