Answers are grounded in this paper and cite their sources.
• We proposed MK-AcroFOD for cross-domain few-shot PWD detection, solving sample scarcity without significant modifications to the model. • CPCA attention and MPDIoU loss are integrated to boost feature extraction and bounding box accuracy. • Our model achieved high-precision detection with mAP@0.5 of 93.44% and Recall of 90.00% under limited cross-domain samples, and can be extended to other forest disease detections. Pine forests around the globe face a major threat from pine wilt disease (PWD). Based on the dataset obtained in Weihai City, we proposed a model named MultiSight Pine Wilt Few-Shot Object Detection (MS-PWFOD). The model is combined with the Adaptive Method for Cross-domain Few-shot Object Detection (AcroFOD) detection method, which enhances its capability to adapt to various environmental situations with a relatively limited quantity of samples. We also replace the original Maximum Mean Discrepancy (MMD) function with Multi-kernel MK-MMD, which can more effectively capture features and complex structures, thereby enhancing the generalization ability. MS-PWFOD also incorporates Channel Prior Convolutional Attention (CPCA) to focus on crucial image features. The Minimum Point Distance Intersection over Union (MPDIoU) is also included to decrease computational demands, while enhancing the precision. Finally, this new MS-PWFOD model achieves excellent detection performances with Precision, Recall, mAP@0.5, and F1-score in high values of 85.71 %, 90.00 %, 93.44 %, and 87.80 %, respectively. Our study improves the generalization and precision of the PWD detection model when it is difficult to obtain the training dataset, thus promoting the maintenance and management of pine forests.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: