Journal of Engineering, Project, and Production Management, 2026, 16(4), 2026-0007
DAMF-YOLOv8: A Multi-Class Apple Detection Algorithm Based on Deformable Attention and Multi-Scale
Undergraduate Student, School of Automation Science and Electrical Engineering, Beihang University, 37 Xueyuan Road, Haidian District, Beijing, 100191, P.R. China, E-mail: 23376046@buaa.edu.cn
Project Management
Received April 14, 2026; revised May 14, 2026; accepted May 14, 2026
Available online June 4, 2026
Abstract: Automated apple detection in orchard environments presents significant challenges due to the prevalence of small targets, frequent occlusions, and the need for multi-category classification under complex conditions. To navigate toward solutions, this study introduces Deformable Attention and Multi-scale Fusion YOLOv8 (DAMF-YOLOv8), an enhanced detection model based on YOLOv8 that incorporates three existing modules. The first is the C2f-DLKA module, which combines deformable convolution with large-kernel attention to improve the extraction of small-target features and contextual modeling. The second is the Multi-Scale Local Channel Attention (MLCA) mechanism, employing multi-scale local-global channel attention to achieve adaptive fusion of spatial and semantic features. The third is the Detect-AFPN-P2345 network, facilitating efficient multi-scale feature fusion across P2 to P5 levels. Through systematic parameter optimization and ablation studies, the integration of these modules is empirically validated, achieving an optimal balance between detection accuracy and model efficiency. Experimental results demonstrate that DAMF-YOLOv8 achieves 91.8% mAP50 and 72.4% mAP50-95, representing improvements of 1.8 and 2.0 percentage points, respectively, over the baseline while maintaining minimal parameter growth. For challenging samples such as green apples, the model improves mAP50 by 4.6% and mAP50-95 by 3.9%. By integrating these modules, DAMF-YOLOv8 offers an effective, lightweight solution for automated orchard monitoring.
Keywords: Apple detection; lightweight model; attention mechanism; feature pyramid network; precision agriculture.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Meng, Z. (2026). DAMF-YOLOv8: A Multi-Class Apple Detection Algorithm Based on Deformable Attention and Multi-Scale. Journal of Engineering, Project, and Production Management, 16(4), 2026-0007.
