"Sweet Potato Virus Disease (SPVD) Datasets"

"The dataset for SPVD was collected from experimental fields located in Xuzhou, China, focusing on the widely cultivated cultivar \"Xushu 3\". Data acquisition was conducted on July 24-25, 2024, using both handheld RGB cameras (Nikon D7100) and portable video recording devices (DJI Pocket 2).After image collection, manual annotation was conducted by experienced plant pathologists to ensure high-quality labeling of SPVD symptoms. Each image was evaluated at the individual plant level, and labels were assigned based on visible symptom characteristics as defined by agronomic and virological standards. The initial dataset consists of 1,264 original images, which were labeled into several symptomatic categories including Healthy, Leaf Chlorosis, Mild, Moderate, Narrow Leaves, SPVD, Serious, Vein Distortion, and Vein Yellowing. To address the imbalance and improve the diversity of the dataset, 1,093 enhanced samples were generated using advanced data augmentation techniques such as rotation, contrast adjustment, horizontal\/vertical flipping, and random cropping.To ensure reproducibility and reliability of the SPVD image dataset, a detailed experimental protocol was followed. RGB images were captured using a Nikon D7100 DSLR at a resolution of 6000\u00d74000 pixels and a DJI Pocket 2 at 3840\u00d72160 pixels. Handheld data were acquired from three height ranges (0.5\u20131.5 m) and three viewing angles (0\u00b0, 30\u00b0, and 60\u00b0) per plant to enrich geometric diversity. The field layout followed a randomized block design covering, with manual symptom scoring for cross-reference. All annotations were performed using the LabelImg tool, adhering to predefined disease symptom ontology validated by plant pathologists."