Project-AgML/pdsnet_field_plant_disease_segmentation

Pdsnet Field Plant Disease Segmentation

Field plant disease segmentation images for three diseases, 80 images each: cedar apple rust on apple, bacterial blight on soybean, and stripe rust on wheat. The wheat images were taken in the field at the Jiaozuo Academy of Agriculture and Forestry Sciences (Henan, China) in May 2023; the soybean and apple images come from public sources. Each image was padded to a square with black pixels and resized to 640 x 640, then labelled pixel by pixel into three classes: 0 background, 1 healthy leaf and 2 disease spots, so disease severity can be computed as spots / (spots + healthy leaf). The disease and crop columns record each image's disease. The dataset contains 240 images with pixel-level mask annotations.

This dataset is indexed on https://project-agml.github.io/ as part of the AgML python library.

Citation

@article{zhao2025plant,
  title={Plant Disease Segmentation Networks for Fast Automatic Severity Estimation Under Natural Field Scenarios},
  author={Zhao, Chenyi and Li, Changchun and Wang, Xin and Wu, Xifang and Du, Yongquan and Chai, Huabin and Cai, Taiyi and Xiang, Hengmao and Jiao, Yinghua},
  journal={Agriculture},
  volume={15},
  pages={583},
  year={2025},
  publisher={MDPI}
}

This dataset was reformatted from its original format to match HuggingFace standards.