---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: mask
    dtype: image
  - name: split
    dtype: string
  splits:
  - name: train
    num_bytes: 29494844
    num_examples: 788
  download_size: 29526452
  dataset_size: 29494844
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
task_categories:
- image-segmentation
size_categories:
- n<1K
---
# Cwf788 Weed Segmentation

This dataset provides real-world RGB imagery for semantic segmentation of weeds in agricultural fields. It captures diverse weed instances under natural field conditions, offering a practical resource for developing and evaluating computer vision models in precision agriculture applications. The dataset contains 788 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{li2019real,
  title={Real-time crop recognition in transplanted fields with prominent weed growth: a visual-attention-based approach},
  author={Li, Nan and Zhang, Xiaoguang and Zhang, Chunlong and Guo, Huiwen and Sun, Zhe and Wu, Xinyu},
  journal={IEEE Access},
  volume={7},
  pages={185310--185321},
  year={2019},
  publisher={IEEE}
}
```

https://github.com/ZhangXG001/Real-Time-Crop-Recognition

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