CNN Optimization Based on Transfer Learning through a Combination of Specialized Augmentation and Fine-Tuning for Organic-Inorganic Waste Classification
Keywords:
Data Augmentation, Fine-Tuning, ResNet-50, Transfer Learning, Waste ClassificationAbstract
Waste management in Indonesia still relies heavily on manual sorting between organic and inorganic waste, which is slow, inconsistent, and prone to error. This study develops a Convolutional Neural Network (CNN) classification model for organic and inorganic waste images using a ResNet-50 backbone with transfer learning, optimized through a combination of two-phase fine-tuning and class-specific data augmentation. The dataset, Waste Classification Data from Kaggle (25,077 images), was re-split by stratified random sampling into training (17,553), validation (3,762), and test (3,762) subsets. Augmentation was tailored per class: the Organic class received both geometric and color/channel transformations to simulate decay-related color changes, while the Inorganic class received only geometric transformations to preserve material-specific color as a discriminative feature. Training proceeded in two stages: feature extraction with a fully frozen backbone, followed by gradual fine-tuning of the conv4 block and above using a reduced learning rate. An ablation comparison showed that a baseline model (no fine-tuning, no augmentation) reached only 78.02% accuracy, whereas the proposed model achieved 96.23% accuracy, 96.29% precision, 96.23% recall, and a 96.23% F1-score, an 18.21-percentage-point accuracy gain over the baseline. The improvement was consistent across both classes, most notably raising Inorganic recall by 20.52 points. These results confirm that combining gradual fine-tuning with class-specific augmentation substantially improves the generalization ability of a ResNet-50-based CNN for organic-inorganic waste classification, yielding performance competitive with prior transfer-learning approaches on similar tasks.
Downloads
References
Alzubaidi, L., Zhang, J., Humaidi, A. J., Dujaili, A. Al, Duan, Y., Shamma, O. Al, Santamaría, J., Fadhel, M. A., Amidie, M. Al, & Farhan, L. (2021). Review of deep learning: concepts, CNN architectures, challenges, applications, future directions. In Journal of Big Data. Springer International Publishing. https://doi.org/10.1186/s40537-021-00444-8
Arifin, F., Habiburrakhman, M., & Gusti, W. R. (2023). Classification of Organic and Inorganic Waste Types Based on Neural Networks. 8(1), 78–85.
Aulia, D. S., Arwoko, H., & Asmawati, E. (2024). Klasifikasi Sampah Rumah Tangga Menggunakan Metode Convolutional Neural Network. 114–120. https://doi.org/10.47002/metik.v8i2.956
Chatterjee, S., Hazra, D., & Byun, Y. (2022). IncepX-Ensemble : Performance Enhancement Based on Data Augmentation and Hybrid Learning for Recycling Transparent PET Bottles. IEEE Access, 10, 52280–52293. https://doi.org/10.1109/ACCESS.2022.3174076
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., & Zhai, X. (2021). An Image Is Worth 16x16 Words: Transformers For Image Recognition At Scale.
Egger, J., Pepe, A., Gsaxner, C., Jin, Y., Li, J., & Kern, R. (2021). Deep learning — a fi rst meta-survey of selected reviews across scientific disciplines, their commonalities, challenges and research impact. 1–83. https://doi.org/10.7717/peerj-cs.773
Eltehewy, R., Abouelfarag, A., & Saleh, S. N. (2023). Efficient Classification of Imbalanced Natural Disasters Data Using Generative Adversarial Networks for Data Augmentation.
Fu, B., Li, S., Wei, J., Li, Q., Wang, Q., & Tu, J. (2021). A Novel Intelligent Garbage Classification System Based on Deep Learning and an Embedded Linux System. XX, 1–13. https://doi.org/10.1109/ACCESS.2021.3114496
Haar, L. V., Elvira, T., & Ochoa, O. (2023). An analysis of explainability methods for convolutional neural networks. Engineering Applications of Artificial Intelligence, 117, 105606. https://doi.org/10.1016/j.engappai.2022.105606
Ibrahim, A. A. M. S., & Tapamo, J. R. (2024). Transfer learning-based approach using new convolutional neural network classifier for steel surface defects classification. Scientific African, 23(January), e02066. https://doi.org/10.1016/j.sciaf.2024.e02066
Jose, J., Mana, S. C., Babu, K. S., Kalaiarasi, G., & Selvi, M. (2025). Enhancing waste classification accuracy with Channel and Spatial Attention-Based Multiblock Convolutional Network. Environmental Monitoring and Assessment, 197(2), 198. https://doi.org/10.1007/s10661-025-13629-y
Lin, K., Zhao, Y., Gao, X., Zhang, M., Zhao, C., Peng, L., & Zhang, Q. (2022). Applying a deep residual network coupling with transfer learning for recyclable waste sorting. Environmental Science and Pollution Research, 91081–91095. https://doi.org/10.1007/s11356-022-22167-w
Malik, M., Sharma, S., Uddin, M., Chen, C., Wu, C., & Soni, P. (2022). Waste Classification for Sustainable Development Using Image Recognition with Deep Learning Neural Network Models. 1–18.
Mubarokh, fahmi wafi. (2024). Image Classification of Organic and Inorganic Waste Using Convolutional Neural Networks. 781–783.
Puspitasari, F. H., Supriyadi, S., & Al-irsyad, M. (2022). Analysis of Organic and Inorganic Waste Management Towards a Green Campus at Universitas Negeri Malang. 44(Ismophs 2021), 68–76.
Razavi, S. (2021). Deep learning , explained : Fundamentals , explainability , and bridgeability to process-based modelling. Environmental Modelling and Software, 144(August), 105159. https://doi.org/10.1016/j.envsoft.2021.105159
Saleem, M. A., Senan, N., Wahid, F., Aamir, M., Samad, A., & Khan, M. (2022). Comparative Analysis of Recent Architecture of Convolutional Neural Network. 2022. https://doi.org/10.1155/2022/7313612
Saleem, M. H., Potgieter, J., & Arif, K. M. (2022). A weight optimization-based transfer learning approach for plant disease detection of New Zealand vegetables. October, 1–22. https://doi.org/10.3389/fpls.2022.1008079
Sapitri, W., Kunang, Y. N., & Yadi, I. Z. (2023). The Impact of Data Augmentation Techniques on the Recognition of Script Images in Deep Learning Models. 8(2), 169–176. https://doi.org/10.15575/join.v8i2.1073
Sathishkumar, V. E., Cho, J., & Subramanian, M. (2023). Forest fire and smoke detection using deep learning ‑ based learning without forgetting. Fire Ecology. https://doi.org/10.1186/s42408-022-00165-0
Shorten, C., & Khoshgoftaar, T. M. (2019). A survey on Image Data Augmentation for Deep Learning. Journal of Big Data. https://doi.org/10.1186/s40537-019-0197-0
Ting-Wei, W., Zhang, H., Wei, P., Fan, L., & Pin-jing, he. (2023). Applications of convolutional neural networks for intelligent waste identification and recycling. Resources, Conservation and Recycling, 190. https://doi.org/10.1016/j.resconrec.2022.106813
Wu, F., & Lin, H. (2022). Effect of transfer learning on the performance of VGGNet-16 and ResNet-50 for the classification of organic and residual waste. 10(October), 1–11. https://doi.org/10.3389/fenvs.2022.1043843
Yong, L., Ma, L., Sun, D., & Id, L. Du. (2023). Application of MobileNetV2 to waste classification. 1–16. https://doi.org/10.1371/journal.pone.0282336
Yoo, J., & Kang, S. (2023). Class-Adaptive Data Augmentation for Image Classification. IEEE Access, 11(March), 26393–26402. https://doi.org/10.1109/ACCESS.2023.3258179
Zhao, Z., Alzubaidi, L., Zhang, J., Duan, Y., & Gu, Y. (2024). A comparison review of transfer learning and self-supervised learning : Definitions, applications, advantages and limitations. Expert Systems With Applications, 242(June 2023), 122807. https://doi.org/10.1016/j.eswa.2023.122807
