Assessing wheat blast resistance by integrating convolutional neural networks on image analysis: scripts and dataset

This dataset contains wheat spike images, metadata, and R and Python scripts used to assess wheat blast severity and classify Pyricularia oryzae Triticum lineage isolates using image analysis and convolutional neural networks. The repository includes three structured datasets for model training and testing: a general disease severity dataset, an isolate classification dataset, and an isolate-specific severity dataset. It also provides scripts for image preprocessing, including spike segmentation, diseased area segmentation, severity estimation, image augmentation, and CNN training, testing, and prediction. Pre-trained YOLO11n model weights are included for each dataset. In addition, the repository contains an independent image set acquired in 2025 for model prediction and validation, as well as a metadata file describing image number, isolate, cultivar, replicate, spike, image rotation, disease severity, and class assignment. These files support the reproducibility of the analyses presented in the associated study on image-based phenotyping for wheat blast resistance assessment.