Answers are grounded in this paper and cite their sources.
Precise forecasting and efficient loss assessment constitute the key technical prerequisites for agricultural insurance companies to conduct disaster claims verification. This study proposes an improved YOLOKIWI detection model. This model adopts YOLOv11n as its baseline architecture, introducing three core improvement mechanisms: replacing the C3k2 layer in YOLOv11n's backbone network with StarBlock to effectively address occlusion issues caused by foliage in kiwifruit orchards; Employing P2 ultra-shallow layer detection to extend the YOLOv11n multi-class detection head architecture, thereby acquiring high-resolution details to compensate for edge blurring caused by occlusion. This enhances the shallow layer features' sensitivity to local details and improves the model's robustness towards lesions of varying scales; Replacing YOLOv11n's CIoU loss with EIoU loss for bounding box regression optimization enables the model to better mitigate the detection bias towards non-square irregular objects inherent in traditional loss functions. The study findings indicate that YOLOKIWI achieved a mean average precision (mAP) of 90.9%, an accuracy rate of 94.1%, and a recall rate of 89.0% on the self-constructed kiwifruit sunburn dataset. It outperformed mainstream models such as YOLOv8n, YOLOv9t, YOLOv10n, DETR-ResNet50, and Faster R-CNN. Compared to the baseline model YOLOv11n, mAP, recall, and accuracy improved by 3.0%, 6.6%, and 3.4% respectively. The research findings demonstrate that YOLOKIWI achieves stable and reliable detection results in complex backgrounds, rendering it a viable model approach for rapid damage assessment of sunburn in kiwifruit.
No takes yet. Share an insight, caveat, or question.
Niu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: