---
license: cc-by-4.0
task_categories:
- image-segmentation
size_categories:
- n<1K
dataset_info:
  features:
  - name: file_name
    dtype: image
  - name: split
    dtype: string
  - name: mask
    dtype: image
  - name: mask_binary
    dtype: image
  - name: detected_vegetation_image
    dtype: image
  - name: campaign
    dtype: string
  splits:
  - name: train
    num_bytes: 726848394
    num_examples: 160
  download_size: 726881494
  dataset_size: 726848394
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
# Sesame Aerial Weed Segmentation

This dataset provides real aerial RGB imagery of tobacco and sesame crop fields in Pakistan, captured using a drone-mounted camera for weed segmentation tasks. The images were collected in field environments across multiple locations, offering diverse examples of weed distribution patterns in agricultural settings. The dataset contains 160 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

```bibtex
@article{moazzam2023w,
  title={A W-shaped convolutional network for robust crop and weed classification in agriculture},
  author={Moazzam, Syed Imran and Nawaz, Tahir and Qureshi, Waqar S. and Khan, Umar S. and Tiwana, Mohsin Islam},
  journal={Precision Agriculture},
  volume={24},
  pages={2002-2018},
  year={2023},
  publisher={Springer US}
}
```

Moazzam, Imran (2023), “SeSame / Weed Aerial Dataset”, Mendeley Data, V2, doi: 10.17632/9pgv3ktk33.2

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