Class Imbalance Alleviation in Object Detection via YOLOv11-based Deep Dynamic Feature Fusion

Qi  Yang; Bingkun  Jiang; Jiatong  Tang; Jianxi  Huang; Minghao  Li

doi:10.6180/jase.202608_31.019

Class Imbalance Alleviation in Object Detection via YOLOv11-based Deep Dynamic Feature Fusion

Computer Science and Information Engineering

Qi Yang¹, Bingkun Jiang¹, Jiatong Tang¹, Jianxi Huang², and Minghao Li¹This email address is being protected from spambots. You need JavaScript enabled to view it.

¹Shenyang Ligong University

²Fuzhou University

Received: December 31, 2025
Accepted: January 12, 2026
Publication Date: February 14, 2026

Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Download Citation: ||https://doi.org/10.6180/jase.202608_31.019

Object detection models often face significant performance limitations. These challenges include severe sample imbalance, background interference, and target occlusion. Such issues are particularly prevalent in complex industrial and medical imaging domains. While existing solutions typically focus on data resampling or loss function re-weighting to handle imbalance, a fundamental bottleneck within the network architecture itself is often overlooked. Traditional feature fusion necks, such as FPN and PANet, rely on static convolutions that inevitably become biased towards the majority class during training, leading to the marginalization or loss of minority-class features. To address this critical issue at the feature-fusion level, we propose the Semantic-Aware Fusion Neck (SAF-Neck), which replaces the static fusion paradigm with a dynamic, input adaptive mechanism. By generating content-aware convolutional kernels for each input, SAF-Neck adaptively enhances the discriminative features of minority-class samples, preventing them from being suppressed by the majority class. We integrate this core innovation into a synergistic architecture with a Lightweight Probabilistic Spatial Attention-HGNetv2(LPSA-HGNetv2) and an imbalance-robust loss function , forming a comprehensive "front-end feature enhancement and back-end optimization" pipeline. We validate our model, SAF-YOLOv11, on a highly challenging industrial task of coal and gangue classification, characterized by a severe class imbalance ratio of up to 1 : 22. Experimental results show that our model achieves a 90.4% F1-score with a computational load of only 5.7 GFLOPs, outperforming the baseline by 4.1% in F1-score while being 13.6% more computationally efficient.

Keywords: Coal And Gangue Detection; Semantic-Aware Fusion; Lightweight Network; Class Imbalance;YOLOv11n

[1] J.LiandJ.Wang,(2019)“Comprehensiveutilizationand environmental risks of coal gangue: A review "Journal of Cleaner Production 239:117946.
[2] G. Litjens, T. Kooi, B. E. Bejnordi, A. A. A. Setio, F. Ciompi, M. Ghafoorian, J. A. Van Der Laak, B. Van Ginneken, andC. I. Sánchez, (2017)“A survey on deep learning in medical image analysis "Medicalimage analysis42:60–88.
[3] H. Jin,C.Yu,Z.Gong,R.Zheng,Y.Zhao,andQ. Fu,(2023)“Machine learning techniques for pulmonary no dule computer-aided diagnosis using CT images: Asystematic review "Biomedical Signal Processing and Control79:104104.
[4] R. U. Modi, M. Kancheti, A. Subeesh, C. Raj,A.K. Singh, N. S. Chandel, A. S. Dhimate, M. K. Singh, andS.Singh,(2023)“An automated weed identification framework for sugarcanecrop: a deep learning approach" CropProtection173:106360.
[5] A. Gupta, A. Anpalagan, L. Guan, and A. S. Khwaja, (2021)“Deep learning for object detection and sceneperception in selfdriving cars: Survey, challenges, and open issues"Array10:100057.
[6] R. Ameri, C. -C. Hsu, and S. S. Band, (2024) “A system aticreview of deep learning approaches for surface defect detection in industrial applications "Engineering Applications of Artificial Intelligence 130: 107717.
[7] Y.Gao,J.Lin,J.Xie,andZ.Ning,(2020)“Areal-time defect detection method for digital signal processing of industrial inspection applications "IEEE Transactions on Industrial Informatics17(5): 3450–3459.
[8] J. Redmon and A. Farhadi,(2018)“Yolov3:Anincre mental improvement "arXiv preprintarXiv:1804.02767:
[9] A. Bochkovskiy, C. -Y. Wang, and H. -Y. M. Liao,(2020) “Yolov4: Optimal speed and accuracy of object detection" arXivpreprintarXiv:2004.10934:
[10] C. Li, L. Li, H. Jiang, K. Weng, Y. Geng, L. Li, Z. Ke, Q. Li, M. Cheng, W. Nie, etal.,(2022)“YOLOv6:A single-stage object detection frame work for industrial applications "arXivpreprintarXiv: 2209.02976:
[11] A. Wang, H. Chen, L. Liu, K. Chen, Z. Lin, J. Han, etal.,(2024)“Yolov10:Real-timeend-to-endobjectdetection"Advances in Neural Information Processing Systems37:107984–108011.
[12] Z. Lv, Y. Fan, T. Sha, Y. Cui, Y. Wu, H. Lv, M. Sun, Y. Tu, Z. Xu, and W. Wang,(2025)“Alarge-scaleopen imaged at a set for deep learning-enabled intelligent sorting andanalyzingofrawcoal"ScientificData12(1):403. DOI:10.1038/s41597-025-04719-0.
[13] R. Li, L. Zhao, H. Wei, G. Hu, Y. Xu, B. Ouyang, and J.Tan, (2025)“Multi-defect type beam bridged at a set: GYU-DET"ScientificData12(1):1101.DOI:10.1038/s41597-025-05395-w.
[14] Z. Liu, Y. Wang, L. Ma, Y. Wu, G. He, X. Liang, and F. Wang,(2025)“CUs-YOLO:enhancedfeaturefusion model for coal and gangue recognition in complex environment of coal mine "Measurement Science and Technology36(6):065012.
[15] X.WEI,F.WANG,D.HE,C.LIU,andD.XU,(2024) “Coalgangue image recognition model based on CSP Net YOLOv7 target detection algorithm "Coal Science and Technology52(S1): 238–248.
[16] N. Li, K. Qin, X. Li, andA.Zhang,(2025)“AYOLOv7 based coal and gangue recognition model integrating super-resolution reconstruction" Computer Engineering and Applications 61(15): 343–352.
[17] R. Khanam and M. Hussain, (2024) “Yolov11: An overview of the key architectural enhancements" arXiv preprint arXiv:2410.17725:
[18] X. Zhao, W. Zhang, H. Zhang, C. Zheng, J. Ma, and Z. Zhang, (2024) “ITD-YOLOv8: An infrared target detection model based on YOLOv8 for unmanned aerial vehicles" Drones 8(4): 161.
[19] Y. Ge, Z. Li, and L. Meng, (2025) “YOLO-MSD: a robust industrial surface defect detection model via multi scale feature fusion" Applied Intelligence 55(12): 1 18.
[20] W. Lv, S. Xu, Y. Zhao, G. Wang, J. Wei, C. Cui, Y. Du, Q. Dang, and Y. Liu, (2023) “Detrs beat yolos on real-time object detection" CoRR:
[21] H. Shen, Z. Wang, J. Zhang, and M. Zhang, (2024) “L-Net: A lightweight convolutional neural network for de vices with low computing power" Information Sciences 660: 120131.
[22] T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie. “Feature pyramid networks for ob ject detection”. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2017, 2117 2125.
[23] K. Han, Y. Wang, J. Guo, and E. Wu, (2023) “Param eterNet: Parameters are all you need" arXiv preprint arXiv:2306.14525:
[24] Z. Yu, H. Huang, W. Chen, Y. Su, Y. Liu, and X. Wang, (2024) “Yolo-facev2: A scale and occlusion aware face detector" Pattern Recognition 155: 110714.
[25] Z. Cao, L. Fang, Z. Li, and J. Li, (2023) “Lightweight target detection for coal and gangue based on improved Yolov5s" Processes 11(4): 1268.
[26] Y. Sui, L. Zhang, Z. Sun, W. Yi, and M. Wang, (2024) “Research on Coal and Gangue Recognition Based on the Improved YOLOv7-Tiny Target Detection Algorithm" Sensors 24(2): 456.
[27] D. Shang, Z. Lv, Z. Gao, and Y. Li, (2025) “Detection of coal gangue by YOLO deep learning method based on channel pruning" International Journal of Coal Preparation and Utilization 45(1): 231–243.
[28] H. Zhang and K. Ogasawara, (2023) “Grad-CAM based explainable artificial intelligence related to medical text processing" Bioengineering 10(9): 1070.