Project-AgML/WeedsGalore_maize_segmentation

Weedsgalore Maize Segmentation

This dataset provides real-world RGB imagery captured in maize agricultural fields, focusing on weed presence within crop canopies. It includes pixel-level segmentation masks to delineate weeds from maize plants, supporting computer vision research in precision agriculture applications. The dataset contains 780 images with pixel-level mask annotations.

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

The original train/test/val split has been preserved in the split column.

Citation

@inproceedings{celikkan2025weedsgalore,
  title={WeedsGalore: A multispectral and multitemporal UAV-based dataset for crop and weed segmentation in agricultural maize fields},
  author={Celikkan, Ekin and Kunzmann, Timo and Yeskaliyev, Yertay and Itzerott, Sibylle and Klein, Nadja and Herold, Martin},
  booktitle={2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  pages={4767--4777},
  year={2025},
  organization={IEEE}
}

https://github.com/GFZ/weedsgalore

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