Efficient detection of cacao pod diseases using SSD MobileNetV2 FPN-Lite

Detecting and managing cacao pod diseases is an important task for improving crop yield and quality, especially in regions where agriculture serves as a primary livelihood. In this paper we introduce an object detection model based on the SSD MobileNetV2 FPN-Lite architecture for efficient and accurate detection of cacao pod diseases, focusing on “monilla” and “phytophtora”. The model used a feature pyramid network (FPN) to enhance multi-scale detection capabilities to enable the identification of small objects such as early-stage disease symptoms. Evaluation metrics, including mAP, box loss, and classification loss, were used to assess the model's performance. The proposed framework achieved an mAP of 0.83, demonstrating its effectiveness in detecting various cacao pod diseases. With its low computational overhead, the model is optimized for deployment on edge devices, making it a viable solution for real-time disease monitoring in agricultural settings.