Datasets:

Project-AgML
/

grapevine_bunch_condition_detection

dataset_info:
  features:
    - name: image
      dtype: image
    - name: objects
      struct:
        - name: bbox
          list:
            list: float64
        - name: categories
          list:
            class_label:
              names:
                '0': OptimalBunch
                '1': DamagedBunch
    - name: split
      dtype: string
  splits:
    - name: train
      num_bytes: 672763568
      num_examples: 10010
  download_size: 672934058
  dataset_size: 672763568
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - object-detection
size_categories:
  - 10K<n<100K

Grapevine Bunch Condition Detection

This dataset provides real-world RGB images of grapevine bunches captured in a field environment in Vairão, Portugal, using a handheld Xiaomi Redmi Note 7 smartphone. It features high-resolution imagery collected under natural agricultural conditions for object detection tasks focused on grape cluster condition assessment. The images were acquired with a dual-camera smartphone setup, offering practical examples of field-based agricultural imaging for computer vision research. The dataset contains 10,010 images with 11,672 bounding box annotations across 2 categories.

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{pinheiro2023deep,
  title={Deep Learning YOLO-Based Solution for Grape Bunch Detection and Assessment of Biophysical Lesions},
  author={Pinheiro, Isabel and Moreira, Germano and Queirós da Silva, Daniel and Magalhães, Sandro and Valente, António and Moura Oliveira, Paulo and Cunha, Mário and Santos, Filipe},
  journal={Agronomy},
  volume={13},
  pages={1120},
  year={2023},
  publisher={MDPI}
}

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