Abstract:
The detection and classification of plant diseases are critical for improving crop yields
and ensuring food security. This study focuses on developing an enhanced object
detection model tailored for Ash Gourd leaf diseases, a crop prone to significant yield
losses due to common diseases such as Aphid infestation, Downy Mildew, Leaf Curl, and
Leaf Miner. We proposed a customized YOLOv11 architecture with significant
modifications, including the integration of the Convolutional Block Attention Module
(CBAM), an Enhanced Spatial Pyramid Pooling-Fast (SPPF) module, the Hard-Swish
activation function, and an optimized Non-Maximum Suppression (NMS) technique.
These enhancements aimed to improve the model's ability to extract disease-relevant
features while maintaining computational efficiency. The model was trained on a newly
curated dataset of Ash Gourd leaf images, captured under real-world conditions, and
classified into five categories: Healthy, Aphid, Downy Mildew, Leaf Curl, and Leaf
Miner. Experimental results demonstrate that the proposed model achieves superior
performance compared to baseline architectures, with a precision of 89.3%, recall of
88.8%, and mAP@0.5 of 89.3%. Incorporating attention mechanisms and feature fusion
techniques significantly improved detection accuracy, particularly for subtle and
overlapping disease patterns. This study highlights the potential of the customized
YOLOv11 model as a robust and efficient tool for precision agriculture applications.
