A COMPARATIVE STUDY OF OBJECT DETECTION MODELS FOR COFFEE LEAF DISEASE DETECTION
Keywords
DOI:
https://doi.org/10.34238/tnu-jst.15892Abstract
Coffee leaf diseases and pests can reduce crop productivity and quality if they are not detected at an early stage. This study evaluates lightweight object detection models for detecting four categories in the BRACOL dataset, including Cercospora, Miner, Phoma, and Rust. Before model training, the YOLO-format annotations were checked using an automated quality control procedure to verify image-label consistency, annotation format, bounding-box validity, and suspicious bounding boxes. Four YOLO-based baselines, namely YOLOv8n, YOLO11n, YOLOv8s, and YOLO11s, were trained and evaluated using three independent random seeds. In addition, a YOLO11n-MPDIoU variant was investigated to assess whether modifying the bounding-box regression loss could improve localization accuracy without increasing model size. The results show that YOLO11n-MPDIoU obtained the highest average mAP50-95 of 0.5295 ± 0.0190 while maintaining a compact model size of 5.1939 MB. YOLO11s achieved a comparable mAP50-95 of 0.5254 ± 0.0100, but its model size increased to 18.2656 MB. The per-class results indicate that Miner and Phoma were detected more accurately than Cercospora and Rust. Overall, the findings suggest that loss-level localization refinement may provide a modest improvement for lightweight coffee leaf disease detection under the current dataset setting, although the gain over the original YOLO11n was small and further validation on larger and more diverse field datasets is still needed.Downloads
References
J. Jepkoech, D. M. Mugo, B. K. Kenduiywo, and E. C. Too, “Arabica coffee leaf images dataset for coffee leaf disease detection and classification,” Data in Brief, vol. 36, 2021, doi: 10.1016/j.dib.2021.107142.
J. Dantas et al., “A comprehensive review of the coffee leaf miner Leucoptera coffeella (Lepidoptera: Lyonetiidae) - a major pest for the coffee crop in Brazil and others neotropical countries,” Insects, vol. 12, no. 12, 2021, doi: 10.3390/insects12121130.
A. Majeed, F. Faridoon, M. Irfan, and A. Rashid, “Deep learning approaches for precision agriculture: weed and cotton crop classification using YOLO and faster RCNN,” in Proceedings of the 2024 International Conference on IT and Industrial Technologies (ICIT), 2024, pp. 1-6.
A. Vijayakumar and S. Vairavasundaram, “YOLO-based object detection models: a review and its applications,” Multimedia Tools and Applications, vol. 83, pp. 83535-83574, 2024.
A. T. Khan, S. M. Jensen, and A. R. Khan, “Advancing precision agriculture: A comparative analysis of YOLOv8 for multi-class weed detection in cotton cultivation,” Artificial Intelligence in Agriculture, vol. 15, no. 2, pp. 182-191, 2025.
L. Gao, H. Cao, H. Zou, and H. Wu, “DMN-YOLO: a robust YOLOv11 model for detecting apple leaf diseases in complex field conditions,” Agriculture, vol. 15, no. 11, 2025, Art. no. 1138.
A. Sharma, V. Kumar, and L. Longchamps, “Comparative performance of YOLOv8, YOLOv9, YOLOv10, YOLOv11 and Faster R-CNN models for detection of multiple weed species,” Smart Agricultural Technology, vol. 9, 2024, Art. no. 100648.
S. Yu and B. Xiang, “Enhancing rice disease detection through optimized faster R-CNN architecture,” Scientific Bulletin, Series C, vol. 87, no. 4, pp. 110-125, 2025.
J. Fragoso, C. Silva, T. Paixão, A. B. Alvarez, O. C. Júnior, R. Florez, F. Palomino-Quispe, L. G. Savian, and P. A. Trazzi, “Coffee-leaf diseases and pests detection based on YOLO Models,” Applied Sciences, vol. 15, no. 9, 2025, Art. no. 5040.
R. A. Krohling, G. J. M. Esgario, and J. A. Ventura, “BRACOL: A Brazilian Arabica Coffee Leaf Images Dataset for Identification and Quantification of Coffee Diseases and Pests,” Mendeley Data, V1, 2019, doi: 10.17632/yy2k5y8mxg.1.
S. Ma and Y. Xu, “MPDIoU: A loss for efficient and accurate bounding box regression,” arXiv:2307.07662, 2023, doi: 10.48550/arXiv.2307.07662.
A. Abulizi, J. Ye, H. Abudukelimu, and W. Guo, “DM-YOLO: Improved YOLOv9 model for tomato leaf disease detection,” Frontiers in Plant Science, vol. 15, 2025, Art. no. 1473928.
S. Sonawane and N. N. Patil, “Comparative performance analysis of YOLO object detection algorithms for weed detection in agriculture,” Intelligent Decision Technologies, vol. 19, no. 1, pp. 1-12, 2024.
C. Li, Y. Li, X. Chen and Y. Zhang, “Concerning Imbalance and bounding box loss to detect small targets in remote sensing,” IEEE Sensors Journal, vol. 24, no. 17, pp. 27631-27639, 2024.
R. Jain, S. Zaware, N. Kacholia, H. Bhalala, and O. Jagtap, “Advancing underwater trash detection: harnessing mask R-CNN, YOLOv8, EfficientDet-D0 and YOLACT,” 2nd International Conference on Sustainable Computing and Smart Systems (ICSCSS), 2024, pp. 1314-1325.
Z. Yang and Y. Liu, “A steel surface defect detection method based on improved RetinaNet,” Scientific Reports, vol. 15, 2025, Art. no. 6045.
K. Song et al., “Research on high-precision target detection technology for tomato-picking robots in sustainable agriculture,” Sustainability, vol. 17, no. 7, 2025, Art. no. 2885.
M. Huang, Y. Liu, L. Zhao, et al., “A lightweight deep neural network model and its applications based on channel pruning and group vector quantization,” Neural Comput. & Applic., vol. 36, pp. 5333-5346, 2024.
