Coffee Leaf Disease Detection using YOLOv8s-P2-CBAM (Convolutional Block Attention Module) for Small Object Detection

Gaisani, Naila Jinan (2029) Coffee Leaf Disease Detection using YOLOv8s-P2-CBAM (Convolutional Block Attention Module) for Small Object Detection. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22081010061.-cover.pdf Download (1MB)
	Text (BAB I) 22081010061.-bab1.pdf Download (242kB)
	Text (BAB II) 22081010061.-bab2.pdf Restricted to Repository staff only until 13 July 2029. Download (1MB)
	Text (BAB III) 22081010061.-bab3.pdf Restricted to Repository staff only until 13 July 2029. Download (905kB)
	Text (BAB IV) 22081010061.-bab4.pdf Restricted to Repository staff only until 13 July 2029. Download (3MB)
	Text (BAB V) 22081010061.-bab5.pdf Download (230kB)
	Text (Daftar pustaka) 22081010061.-daftarpustaka.pdf Download (196kB)
	Text (LAMPIRAN) 22081010061.-lampiran.pdf Restricted to Repository staff only until 13 July 2029. Download (547kB)

Abstract

Coffee leaf diseases can significantly reduce coffee productivity, making early and accurate disease detection essential. However, detecting small disease lesions remains challenging because they are easily overlooked during feature extraction. This study proposes a modified YOLOv8s model, namely YOLOv8s-P2-CBAM, by integrating a high-resolution P2 detection head and the Convolutional Block Attention Module (CBAM) to improve small object detection performance. The proposed model was trained using the enhanced BRACOL dataset with annotation refinement, data augmentation, and class balancing during the preprocessing stage. Hyperparameter tuning was conducted to determine the optimal training configuration. Model performance was evaluated using Precision, Recall, mAP@50, mAP@50-95, F1-score, confusion matrix analysis, and qualitative visualization. The experimental results show that the proposed YOLOv8s-P2-CBAM model outperformed the standard YOLOv8s model in detecting small coffee leaf disease lesions. The best model achieved a Precision of 0.633, Recall of 0.648, mAP@50 of 0.646, mAP@50-95 of 0.353, and an F1-score of 0.640. These results indicate that the proposed architecture improves localization performance and provides a reliable approach for coffee leaf disease detection, particularly for small lesions.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Puspaningrum, Eva Yulia	NIDN0005078908	evapuspaningrum.if@upnjatim.ac.id
Thesis advisor	Mulyo, Budi Mukhamad	NIDN0718118904	budi.m.mulyo.fasilkom@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Naila Jinan Gaisani

Date Deposited:

13 Jul 2026 04:05

Last Modified:

13 Jul 2026 04:05

URI:

https://repository.upnjatim.ac.id/id/eprint/55182