OPTIMASI YOLOv8n MENGGUNAKAN CONVOLUTIONAL BLOCK ATTENTION MODULE (CBAM) UNTUK DETEKSI PENYAKIT DAUN TANAMAN KEDELAI BERBASIS CITRA DIGITAL

Fikri Alfadani, NIM.: 24206052007 (2026) OPTIMASI YOLOv8n MENGGUNAKAN CONVOLUTIONAL BLOCK ATTENTION MODULE (CBAM) UNTUK DETEKSI PENYAKIT DAUN TANAMAN KEDELAI BERBASIS CITRA DIGITAL. Masters thesis, UIN SUNAN KALIJAGA YOGYAKARTA.

Preview	Text (OPTIMASI YOLOv8n MENGGUNAKAN CONVOLUTIONAL BLOCK ATTENTION MODULE (CBAM) UNTUK DETEKSI PENYAKIT DAUN TANAMAN KEDELAI BERBASIS CITRA DIGITAL) 24206052007_BAB-I_IV-atau-V_DAFTAR-PUSTAKA.pdf - Published Version Download (7MB) | Preview
	Text (OPTIMASI YOLOv8n MENGGUNAKAN CONVOLUTIONAL BLOCK ATTENTION MODULE (CBAM) UNTUK DETEKSI PENYAKIT DAUN TANAMAN KEDELAI BERBASIS CITRA DIGITAL) 24206052007_BAB-II_sampai_SEBELUM-BAB-TERAKHIR.pdf - Published Version Restricted to Registered users only Download (14MB) | Request a copy

Abstract

Soybeans are a strategic food commodity, yet their productivity is hampered by leaf diseases that are difficult to identify quickly and accurately in the field. Although YOLOv8n offers high computational efficiency, its standard backbone lacks an attention mechanism to emphasize relevant spatial and channel features. This limitation can affect detection capabilities in images where disease symptoms are small or share similar visual characteristics. This study aims to evaluate the contribution of the Convolutional Block Attention Module (CBAM) to YOLOv8n's performance through controlled experiments designed to isolate the impact of the CBAM integration. Both models were trained using identical datasets, data splits, and training configurations. A public classification dataset was converted into an object detection format using a transparently documented pseudo-bounding box approach. The performance of the baseline YOLOv8n and YOLOv8n+CBAM was then compared using precision, recall, mAP@50, and mAP@50–95 metrics on the same test data. This approach allowed for a controlled analysis of performance differences attributable to the implemented architectural changes. Test results indicate that YOLOv8n+CBAM outperformed the baseline YOLOv8n across all evaluation metrics. Analysis of the training curves reveals that while both models improved during training, YOLOv8n+CBAM achieved high performance earlier and maintained relatively stable results through the end of the training process. These findings demonstrate that integrating CBAM effectively enhances YOLOv8n's ability to detect soybean leaf diseases while facilitating a more optimal learning process compared to the baseline architecture.

Share this knowledge with your friends :

Actions (login required)

View Item