Benchmarking YOLO26 Against YOLOv11 for Minority Class Waste Detection on an Augmented TACO Dataset

Lingga Kurnia Ramadhani; Bajeng Nurul Widyaningrum

doi:10.30871/jaic.v10i4.13712

Authors

Lingga Kurnia Ramadhani Universitas Ivet
Bajeng Nurul Widyaningrum Politeknik Bina Trada Semarang

DOI:

https://doi.org/10.30871/jaic.v10i4.13712

Keywords:

YOLO26, Class Imbalance, Small Object Detection, Waste Classification, TACO Dataset

Abstract

This study benchmarks YOLO26 against YOLOv11 for detecting minority waste categories (Hazardous/B3 and Residue), evaluating whether YOLO26's Progressive Loss Balancing (ProgLoss) and Small-Target-Aware Label Assignment (STAL) mechanisms address class imbalance and small-object detection challenges. Both models were trained under identical conditions (50 epochs, 640×640) on a combined TACO, RecyBat24, and Food Waste Detection dataset (2,326 images, 70:15:15 split), across three scenarios: YOLOv11 baseline, YOLO26 default, and YOLO26 with class weighting and copy-paste augmentation. YOLO26 (default) achieved a marginally higher mAP@0.5:0.95 (0.393 vs 0.385) and modestly faster CPU inference (≈20% faster) than YOLOv11, with near-identical mAP@0.5 across scenarios. B3 performed consistently well (mAP@0.5 ≈ 0.99), and class weighting improved its detection robustness without raising overall mAP. Residue detection remained the weakest across all scenarios (mAP@0.5 0.076–0.085) and worsened under weighting, indicating that ProgLoss and STAL alone do not resolve its structural visual heterogeneity; this weak, stable Residue performance was confirmed reproducible across three additional training runs with different seeds (mAP@0.5:0.95 = 0.046 ± 0.001). These findings partially support the research hypothesis, positioning YOLO26 (default) as a favorable accuracy-efficiency trade-off for automated waste sorting, while Residue detection requires further data enrichment and augmentation strategies beyond architectural improvements alone.

Downloads

Download data is not yet available.

References

[1] Lingga Kurnia Ramadhani and Bajeng Nurul Widyaningrum, "Implementation of YOLO v11 for Image-Based Litter Detection and Classification in Environmental Management Efforts," Journal of Applied Informatics and Computing (JAIC), vol. 9, no. 3, pp. 617-624, 2025, doi: 10.30871/jaic.v9i3.9213.

[2] Q. Zhang, Q. Yang, X. Zhang, Q. Bao, J. Su, and X. Liu, "Waste image classification based on transfer learning and convolutional neural network," Waste Management, vol. 135, pp. 150-157, 2021, doi: 10.1016/j.wasman.2021.08.038.

[3] J. Gunaseelan, S. Sundaram, and B. Mariyappan, "A Design and Implementation Using an Innovative Deep-Learning Algorithm for Garbage Segregation," Sensors, vol. 23, no. 18, 7963, 2023, doi: 10.3390/s23187963.

[4] F. Fotovvatikhah, I. Ahmedy, R. M. Noor, and M. U. Munir, "A Systematic Review of AI-Based Techniques for Automated Waste Classification," Sensors, vol. 25, no. 10, 3181, 2025, doi: 10.3390/s25103181.

[5] P. A. Rajeev, V. Dharewa, D. Lakshmi, G. Vishnuvarthanan, J. Giri, T. Sathish, and M. Alrashoud, "Advancing e-waste classification with customizable YOLO based deep learning models," Scientific Reports, vol. 15, no. 1, 18151, 2025, doi: 10.1038/s41598-025-94772-x.

[6] X. C. Acaro Chacu00f3n, F. Lo Scudo, G. Cappuccino, and C. Dodaro, "RecyBat24: a dataset for detecting lithium-ion batteries in electronic waste disposal," Scientific Data, vol. 12, no. 843, 2025, doi: 10.1038/s41597-025-05211-5.

[7] M. H. Dipo, F. Al Farid, M. S. Al Mahmud, M. Momtaz, S. Rahman, J. Uddin, and H. A. Karim, "Real-Time Waste Detection and Classification Using YOLOv12-Based Deep Learning Model," Digital, vol. 5, no. 2, 19, 2025, doi: 10.3390/digital5020019.

[8] Y. Zhou, L. Lin, and T. Wang, "Garbage classification detection system based on the YOLOv8 algorithm," AIP Advances, vol. 14, no. 12, 125012, 2024, doi: 10.1063/5.0233953.

[9] S. Riyadi, A. D. Andriyani, and A. M. Masyhur, "Classification of Recyclable Waste Using Deep Learning: A Comparison of Yolo Models," Revue d'Intelligence Artificielle, vol. 38, no. 4, pp. 1089-1096, 2024, doi: 10.18280/ria.380404.

[10] K. Oksuz, B. C. Cam, S. Kalkan, and E. Akbas, "Imbalance Problems in Object Detection: A Review," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 10, pp. 3388-3415, 2021, doi: 10.1109/TPAMI.2020.2981890.

[11] Y. Liu, P. Sun, N. Wergeles, and Y. Shang, "A survey and performance evaluation of deep learning methods for small object detection," Expert Systems with Applications, vol. 172, 114602, 2021, doi: 10.1016/j.eswa.2021.114602.

[12] R. Sapkota, R. H. Cheppally, A. Sharda, and M. Karkee, "YOLO26: Key Architectural Enhancements and Performance Benchmarking for Real-Time Object Detection," arXiv:2509.25164, 2026. [Not yet Scopus-indexed — preprint, included as a supporting technical reference]

[13] P. F. Proença and P. Simões, "TACO: Trash Annotations in Context for Litter Detection," arXiv:2003.06975, 2020. [Dataset source, not Scopus-indexed]

[14] G. Ghiasi, Y. Cui, A. Srinivas, R. Qian, T.-Y. Lin, E. D. Cubuk, Q. V. Le, and B. Zoph, "Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation," in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2021, pp. 2918-2928, doi: 10.1109/CVPR46437.2021.00294.

[15] X. Yu, F. Li, Y. Liu, and A. Wang, "A Multi-Stage Adaptive Copy-Paste Data Augmentation Algorithm Based on Model Training Preferences," Electronics, vol. 12, no. 17, 3695, 2023, doi: 10.3390/electronics12173695.