Vision-Based Papaya Leaf and Fruit Detection Using an Attenuation-Enhanced YOLOv5s Network

Abstract

Papaya leaf and fruit detection using the eyes will be of great significance in intelligent agriculture monitoring and yield determination in the real-field environment. In this research, the authors suggest an attenuation-enhanced YOLOv5s network to detect papaya leaves and fruits effectively and efficiently in natural and complex settings. The proposed model adds an attenuation process to the YOLOv5s backbone, which removes redundant features and enhances discriminative spatial representations to enhance detection stability with different illumination, occlusions, and detection with duality. The extensive experiments that have been carried out on a custom papaya image dataset show that the proposed approach has better detection performance with average mAP@0.5, precision, and recall of 98%, 97%, and 98% respectively. The model architecture has 214 layers and 7,025,023 parameters and gradients and has a computational cost of 16.0 GFLOPs, which is enough to verify its appropriateness as a real-time agricultural model. Comparative analysis shows that attenuation-enhanced YOLOv5s network can be used as a dependable solution to automated papaya leaf and fruit detection in precision agriculture systems as it is more accurate and efficient than the baseline detection models.