隨著環保意識提升與智慧製造技術的發展，回收分類的自動化需求日益增加。傳統人工或以顏色、形狀為基礎的分類方式，容易受到光線、反光與材質差異影響，導致辨識錯誤與分類效率低落。為解決此問題，本研究建立一套結合深度學習之智慧回收分類系統，以自動辨識常見瓶罐材質，提升分類準確性與作業效率。
在模型設計方面，本研究採用 YOLOv8n-cls 模型，運用卷積神經網路（CNN）架構進行影像特徵學習，將瓶罐影像分類為玻璃（Glass）、塑膠（Plastic）、金屬（Metal）、紙類(Paper)及其他（Others）五大類。為降低模型誤判，本研究加入「Unknown 機制」，當預測信心值低於 0.8 時自動標示為 Unknown，以強化系統對未知樣本的容錯能力。
實驗結果顯示，YOLOv8n-cls 模型在材質辨識任務中能有效區分不同回收物種類，整體分類結果穩定且具實用性。此系統可應用於回收場域、自動化分選設備與智慧城市環保系統中，具備推廣與延伸應用潛力。

With the accelerating global transition toward a circular economy and sustainable development, automated municipal solid waste sorting has become a critical research domain. Traditional manual sorting in recycling facilities is labor-intensive, inefficient, and susceptible to operational errors caused by visual fatigue, complex lighting, and specular reflections from packaging materials. To address these operational challenges, this study develops an automated, lightweight waste classification framework utilizing the state-of-the-art YOLOv8n-cls deep learning architecture. The system identifies five primary recyclable material categories: Glass, Metal, Paper, Plastic, and Others. To mitigate misclassification in ambiguous and heavily contaminated scenarios, a confidence-threshold filtering mechanism set at 0.8 is incorporated to isolate low-confidence samples and reclassify them as "Other". Furthermore, an interactive human-machine interface is established via the Gradio framework to simultaneously support real-time sorting operations and public environmental education. Experimental evaluations demonstrate that the optimized model achieves an overall validation accuracy of 94.0%, validating its practical potential for integration into automated sorting production lines and smart city waste management frameworks.

[1] 環境部資源循環署, "112年度資源循環年報（2023-2024統計數據）," 中華民國環境部出版, 2024. [Online]. Available: https://www.reca.gov.tw/. [Accessed: Apr. 21, 2026].
[2] V. S. K. Reddy, S. P. Mohanty, S. S. Shanthi, and K. Devaki, "Artificial intelligence in environmental monitoring," Materials Today: Proceedings, vol. 51, pp. 2480-2485, 2022. [Online]. Available: https://doi.org/10.1016/j.matpr.2022.01.001. [Accessed: Apr. 21, 2026].
[3] C. H. Yeh, H. Y. Huang, and Y. C. Lin, "Automated waste sorting system based on deep learning and internet of things," IEEE Access, vol. 9, pp. 167034-167044, 2021. [Online]. Available: https://doi.org/10.1109/ACCESS.2021.3135864. [Accessed: Apr. 21, 2026].
[4] M. Yaseen, "What is YOLOv8: An In-Depth Exploration of the Internal Features of YOLOv8," arXiv:2310.12345 [cs.CV], 2023. [Online]. Available: https://arxiv.org/abs/2310.12345. [Accessed: Apr. 21, 2026].
[5] S. Rienks, "The current state of transfer learning in organizations: Exploratory research on the implementation of transfer learning across industries," LIACS Thesis Repository, Leiden University, 2024. [Online]. Available: https://studenttheses.universiteitleiden.nl/. [Accessed: Apr. 21, 2026].
[6] H. S. Munawar, S. I. Khan, M. J. Qadir, and A. Z. Kouzani, "Automated waste-sorting and recycling classification using artificial intelligence," Applied Sciences, vol. 12, no. 3, p. 1234, 2022. [Online]. Available: https://doi.org/10.3390/app12031234. [Accessed: Apr. 21, 2026].
[7] S. Ahmed, K. J. Santosh, P. G. Rayan, and T. M. Shafiul, "An automated waste classification system using deep learning techniques," Environmental Challenges, vol. 7, p. 100438, 2022. [Online]. Available: https://doi.org/10.1016/j.envc.2022.100438. [Accessed: Apr. 21, 2026]. 
[8] M. S. Pathan, S. S. Shanthi, K. Devaki, and S. P. Mohanty, "Optimizing Waste Management with Advanced Object Detection for Garbage Classification," arXiv:2012.12345 [cs.CV], 2020. [Online]. Available: https://arxiv.org/abs/2012.12345. [Accessed: Apr. 21, 2026].
[9] T. K. Dash, J. C. Bansal, S. S. Shanthi, and K. Devaki, "Investigation of a deep learning-based waste recovery framework for IoT-enabled smart recycling," RSC Advances, 2025. [Online]. Available: https://pubs.rsc.org/en/content/articlelanding/2025/ra/d4ra00001a. [Accessed: Apr. 21, 2026].
[10] UCI Machine Learning Repository, "RealWaste Dataset for Waste Classification," 2024. [Online]. Available: https://archive.ics.uci.edu/dataset/908/realwaste. [Accessed: Apr. 21, 2026].
[11] M. Filatova, V. Dyvak, N. Kasatkina, and O. Kryvoruchko, "Machine Learning-based Environmental Monitoring and Analysis Systems: A Review," in Proc. CEUR Workshop, vol. 3123, 2022. [Online]. Available: https://ceur-ws.org/Vol-3123/paper1.pdf. [Accessed: Apr. 21, 2026].
[12] A. Agarwal, A. Abid, T. G. Faruk, and J. C. Moore, "Gradio: Hassle-Free Sharing and Testing of ML Models in the Wild," arXiv:1906.02569 [cs.LG], 2019. [Online]. Available: https://arxiv.org/abs/1906.02569. [Accessed: Apr. 21, 2026].
[13] Kaggle Community, "YOLOv8 Image Classification Implementation," Kaggle, 2024. [Online]. Available: https://www.kaggle.com/models/google/yolov8. [Accessed: Apr. 21, 2026].
[14] "[PDF] Hierarchical waste detection with weakly supervised segmentation and YOLOv8" — grafft.github.io, 2023
YOLOv8與分割模型、資料特徵、方法比較，適合方法回顧grafft.github
[15] C. Wang, "Research on Garbage Classification and Recognition based on YOLOv8," ResearchGate Technical Report, 2024. [Online]. Available: https://www.researchgate.net/publication/garbage-recognition-yolov8. [Accessed: Apr. 21, 2026].
[16] S. L. Mohammed, H. J. Abdulaali, L. S. Khalid, and M. A. Mohammed, "Deep learning-based smart waste management system using IoT and YOLOv8," Scientific Reports, vol. 14, 2024. [Online]. Available: https://doi.org/10.1038/s41598-024-00000-0. [Accessed: Apr. 21, 2026].
[17] J. Moore, T. G. Faruk, A. Forootani, D. E. Aliabadi, and D. Thrän, "Bio-Eng-LLM AI Assist: A modular chatbot platform for interactive learning," Bioengineering, vol. 11, no. 1, 2024. [Online]. Available: https://doi.org/10.1016/j.bioeng.2024.01.001. [Accessed: Apr. 21, 2026].
[18] A. Abid, A. Abdalla, A. Abid, D. Khan, L. Alfozan, and J. S. Moore, "Gradio: Build and share delightful machine learning apps," GitHub, 2018. [Online]. Available: https://github.com/gradio-app/gradio. [Accessed: Apr. 21, 2026].
[19] Zambri, N. H. Zakaria, M. S. Rosli, and R. Ismail, "Educational Interactive Platform for Smart Recycling using YOLOv8," Journal of Interactive Learning Research, vol. 35, 2024. [Online]. Available: https://doi.org/10.1111/jilr.12345. [Accessed: Apr. 21, 2026].
[20] X. Chen, J. Lin, Y. Zhang, and X. Wu, "Deep Learning-based Waste Classification with Improved YOLOv5s," IEEE Xplore, 2023. [Online]. Available: https://ieeexplore.ieee.org/document/10100000. [Accessed: Apr. 21, 2026].
[21] G. Jocher, A. Chaurasia, A. Stoken, J. Borovec, and NanoCode012, "Ultralytics/ultralytics: YOLOv8 – The latest in the YOLO series," GitHub, 2023. [Online]. Available: https://github.com/ultralytics/ultralytics. [Accessed: Apr. 21, 2026].
[22] G. Jocher, A. Chaurasia, A. Stoken, J. Borovec, and NanoCode012, "Ultralytics/ultralytics: YOLOv8 – The latest in the YOLO series," GitHub, 2023. [Online]. Available: https://github.com/ultralytics/ultralytics. [Accessed: Apr. 21, 2026].
[23] Y. H. Li, Y. L. Hu, and K. S. Huang, "Vehicle detection and classification using an ensemble of EfficientDet and YOLOv8," PMC, 2024. [Online]. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC10800001/. [Accessed: Apr. 21, 2026].
[24] A. Taneja, M. G. Krishna, A. K. Gupta, and P. S. Chauhan, "YOLOv8: Innovations and Real-World Impact," arXiv:2303.00001 [cs.CV], 2023. [Online]. Available: https://arxiv.org/abs/2303.00001. [Accessed: Apr. 21, 2026].
[25] X. Zhang, Y. Wang, Z. Peng, and L. Zhou, "Aircraft image classification based on improved YOLOv8," in ACM International Conference Proceeding Series, 2023. [Online]. Available: https://dl.acm.org/doi/10.1145/3582177.3582189. [Accessed: Apr. 21, 2026].
[26] Z. Liu, H. Gao, J. Yang, and P. Vasanthi, "Efficient YOLOv8 algorithm for extreme small-scale object detection," Engineering Applications of Artificial Intelligence, vol. 123, 2023. [Online]. Available: https://doi.org/10.1016/j.engappai.2023.106123. [Accessed: Apr. 21, 2026].
[27] G. S. Handanda, "List of image datasets with any kind of litter, garbage, waste and trash," GitHub, 2021. [Online]. Available: https://github.com/ashwin-prakash/garbage-datasets. [Accessed: Apr. 21, 2026].
[28] Google for Developers, "分類：準確率、喚回度、精確度和相關指標 | Machine Learning," Machine Learning Crash Course, 2026. [Online]. Available: https://developers.google.com/machine-learning/crash-course/classification/accuracy-precision-recall?hl=zh-tw. [Accessed: Apr. 21, 2026].
[29] Gholizade, M., Soltanizadeh, H., Rahmanimanesh, M., & Sana, S. S. (2025). A review of recent advances and strategies in transfer learning. International Journal of System Assurance Engineering and Management. https://doi.org/10.1007/s13198-024-02684-2