Datasets:

Project-AgML
/

paddy_weed_target_detection

dataset_info:
  features:
    - name: image
      dtype: image
    - name: objects
      struct:
        - name: bbox
          list:
            list: float64
        - name: categories
          list:
            class_label:
              names:
                '0': weed
  splits:
    - name: train
      num_bytes: 12796353
      num_examples: 358
  download_size: 12800061
  dataset_size: 12796353
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - object-detection
size_categories:
  - n<1K

Paddy Weed Target Detection

This dataset provides real-world RGB images captured in rice paddy fields during the seedling stage, focusing on the detection of weeds within rice crops under natural field conditions. Images were collected using a handheld Canon IXUS 1000 HS camera with a telephoto lens, offering close-up field-level perspectives for agricultural computer vision applications. The dataset contains 358 images with 840 bounding box annotations across 1 category.

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

Citation

@article{deng2023weed,
  title={Weed target detection at seedling stage in paddy fields based on YOLOX},
  author={Deng, Xiangwu and Qi, Long and Liu, Zhuwen and Liang, Song and Gong, Kunsong and Qiu, Guangjun},
  journal={PLOS ONE},
  volume={18},
  pages={e0294709},
  year={2023},
  publisher={Public Library of Science}
}

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