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

Description

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.

Steps to reproduce

The repository is organized into three main folders. The code/ folder contains the R and Python scripts for image processing and deep learning, as well as pre-trained YOLO11n model weights. The model_data/ folder contains the structured datasets used for model training and testing. The predict/ folder contains the independent image set acquired in 2025 and the corresponding R and Python scripts used for model prediction. A recommended workflow is: Choose either the R or Python implementation. Use the scripts in code/ to preprocess images, segment spikes and diseased areas, estimate severity, and generate augmented data. Use the datasets in model_data/ to train and test the CNN models. Use the scripts in predict/ to run inference on the independent 2025 image set.

Institutions

• Universidade Federal de Lavras

  Minas Gerais, Lavras
• North Carolina State University

  North Carolina, Raleigh
• Louisiana State University

  Louisiana, Baton Rouge

Categories

Funders

• Coordenação de Aperfeicoamento de Pessoal de Nível Superior

  Ministry of Education

  Brasília
• Fundação de Amparo à Pesquisa do Estado de Minas Gerais

  Federal Government of Brazil

  Belo Horizonte
• Conselho Nacional de Desenvolvimento Científico e Tecnológico

  Ministry of Science, Technology and Innovation

  Brasília

Licence