Datasets:

Project-AgML
/

mtl_plotcounter_soybean_detection

dataset_info:
  features:
    - name: image
      dtype: image
    - name: objects
      struct:
        - name: bbox
          list:
            list: float64
        - name: categories
          list:
            class_label:
              names:
                '0': soybean_seedling
    - name: plot_id
      dtype: int32
    - name: variety
      dtype: string
    - name: seeds_per_hole
      dtype: int32
    - name: sowing_density_seeds_per_ha
      dtype: int32
    - name: seedling_count
      dtype: int32
    - name: split
      dtype: string
  splits:
    - name: train
      num_bytes: 137406490
      num_examples: 54
  download_size: 137397307
  dataset_size: 137406490
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
task_categories:
  - object-detection
size_categories:
  - n<1K

Mtl Plotcounter Soybean Detection

Plot-scale soybean seedling detection from UAV imagery of a 2023 breeding trial in Jinzhong, Shanxi, China. Two varieties (Dongdou 1 and Tiefeng 31) were sown at 1, 2 or 3 seeds per hole and at 120,000, 154,000 or 222,000 seeds per hectare in a factorial design with three replicates, giving 54 plots. A DJI Mavic 2 Pro flown at 5 m (0.1 cm per pixel) imaged the field after emergence; each plot was cut from the mosaic and zero-padded to 3072 x 6144 pixels. Because seedlings overlap heavily in the imagery, every seedling (5,932 in total, 55 to 211 per plot) was marked in the field on printed plot images and then boxed. Each row also carries the plot's variety, seeds per hole, sowing density and seedling count, and the split column keeps the authors' train/val/test assignment. The dataset contains 54 images with 5,932 bounding box annotations across 1 category.

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

@article{xue2025mtl,
  title={MTL-PlotCounter: Multitask Driven Soybean Seedling Counting at the Plot Scale Based on UAV Imagery},
  author={Xue, Xiaoqin and Li, Chenfei and Liu, Zonglin and Sun, Yile and Li, Xuru and Song, Haiyan},
  journal={Remote Sensing},
  volume={17},
  pages={2688},
  year={2025},
  publisher={MDPI}
}

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