Detection algorithm dataset for PWD in complex environments

The data used in this study were primarily collected between February and June 2024 through multiple aerial surveys conducted in two cities in northern and southern China, using the DJI Mini 2 drone developed by DJI Innovations, Shenzhen. The drone was flown at relative altitudes ranging from 50 to 120 meters above ground level, following an "S"-shaped flight path to comprehensively cover areas populated with pine trees.Based on the progression of pine wilt disease symptoms, the collected data were categorized into two classes: "early stage" and "late stage." A total of 2,775 images were compiled and divided into training, validation, and test sets in an 8:1:1 ratio. To expand the dataset, a series of augmentation techniques—including flipping, scaling, noise addition, and cropping—were applied, increasing the total number of images to 18,375.Addressing the challenge of reduced detection performance in complex environments, an improved YOLOv5 model was developed to enhance identification accuracy and support large-scale, rapid monitoring. All data and code used in the analysis have been made publicly available through Mendeley Data to ensure experimental reproducibility. While the training and inference data used in the deep learning experiments exceed 80 GB in size, only a portion of the dataset could be uploaded due to storage limitations on Mendeley Data. Researchers interested in replicating this study are welcome to contact me via email, and I will gladly provide access to the full dataset through alternative repositories.