Datasets:

Project-AgML
/

FruitVision_quality_classification

configs:
  - config_name: raw
    default: true
    data_dir: raw
    features:
      - name: image
        dtype: image
      - name: label
        dtype:
          class_label:
            names:
              '0': Formalin-mixed
              '1': Fresh
              '2': Rotten
      - name: fruit
        dtype: string
  - config_name: augmented
    data_dir: augmented
    features:
      - name: image
        dtype: image
      - name: label
        dtype:
          class_label:
            names:
              '0': Formalin-mixed
              '1': Fresh
              '2': Rotten
      - name: fruit
        dtype: string
license: cc-by-nc-nd-4.0
task_categories:
  - image-classification
size_categories:
  - 10K<n<100K

FruitVision Quality Classification

A dataset for quality classification of apples, bananas, mangoes, grapes, and oranges. The dataset contains raw and augmented versions.
The raw dataset contains 10,154 images.
Images per class:

Formalin-mixed: 3,176
Fresh: 3,800
Rotten: 3,178

The augmented dataset contains 73,389 images.
Images per class:

Formalin-mixed: 22,228
Fresh: 30,400
Rotten: 20,761

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

Citation

@article{bijoy2025fruitvision,
  title={FruitVision: A benchmark dataset for fresh, rotten, and formalin-mixed fruit detection},
  author={Bijoy, Md Hasan Imam and Tasnim, Syeda Zarin and Awsaf, Syed Ali and Hasan, Md Zahid},
  journal={Data in Brief},
  volume={61},
  pages={111752},
  year={2025},
  publisher={Elsevier}
}

Bijoy, Md Hasan Imam; Tasnim, Syeda Zarin; Awsaf, Syed Ali; Hasan, Md Zahid (2025), “FruitVision: A Benchmark Dataset for Fresh, Rotten, and Formalin-mixed Fruit Detection”, Mendeley Data, V2, doi: 10.17632/xkbjx8959c.2