Answers are grounded in this paper and cite their sources.
Timely and accurate detection of plant diseases is essential for improving crop yields and ensuring food security, particularly in regions like Cameroon, where farmers often rely on visual inspection. An approach limited by subjectivity and low precision. Although deep learning and precision agriculture technologies have advanced significantly, many models still face challenges in detecting early-stage symptoms, especially in real-world, resource-limited environments. This study introduces YOLO-AgriNet, a customized object detection model built on the YOLOv8 architecture, optimized for plant disease detection under tropical and low-resource conditions. To enhance the detection of small and subtle features, YOLO-AgriNet integrates key architectural improvements, including Convolutional Block Attention Modules (CBAM), Atrous Spatial Pyramid Pooling (ASPP) and an additional Stage Layer 5 for finer spatial representation. The model was trained on a public dataset and a curated set of local images from plantations in Cameroon. Compared to YOLOv8, Faster R-CNN, and SSD, YOLO-AgriNet achieved higher performance with a mAP@0.5 of 84.5%, real-time inference speed (45 FPS), and improved robustness in complex tropical conditions. It also demonstrated superior accuracy in detecting small disease symptoms and reduced false positives. YOLO-AgriNet provides a lightweight, scalable, and practical solution for real-time plant disease monitoring. Its compatibility with low-cost platforms like smartphones and drones makes it highly suitable for developing regions, enabling timely interventions and supporting sustainable agriculture.
No takes yet. Share an insight, caveat, or question.
Nkonjoh et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: