---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    dtype:
      class_label:
        names:
          '0': aphids
          '1': armyworm
          '2': beetle
          '3': bollworm
          '4': grasshopper
          '5': mites
          '6': mosquito
          '7': sawfly
          '8': stem_borer
  - name: split
    dtype: string
  splits:
  - name: train
    num_bytes: 66288874
    num_examples: 2616
  download_size: 64688955
  dataset_size: 66288874
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-4.0
task_categories:
- image-classification
size_categories:
- 1K<n<10K
---
# Bau Insectv2 Classification

This dataset comprises real-world RGB images of insect species captured in agricultural field environments across Bangladesh. Collected using Samsung smartphone and Canon digital cameras during field surveys, it provides diverse visual samples for insect classification tasks in crop monitoring contexts. The dataset contains 2,616 images across 9 classes: aphids, armyworm, beetle, bollworm, grasshopper, mites, mosquito, sawfly, stem_borer.  
Images per class:
- aphids: 310
- armyworm: 266
- beetle: 341
- bollworm: 281
- grasshopper: 323
- mites: 296
- mosquito: 345
- sawfly: 237
- stem_borer: 217

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{salehin2024bau,
  title={BAU-Insectv2: An agricultural plant insect dataset for deep learning and biomedical image analysis},
  author={Salehin, Imrus and Khan, Mahbubur Rahman and Habiba, Ummya and Badhon, Nazmul Huda and Moon, Nazmun Nessa},
  journal={Data in Brief},
  volume={53},
  pages={110083},
  year={2024},
  publisher={Elsevier}
}
```

Salehin, Imrus; Angon, Prodipto Bishnu ; Habiba, Ummya  (2023), “BAU-Insectv2  Agricultural Plant Insect Dataset ”, Mendeley Data, V2, doi: 10.17632/x2c6c8thdk.2

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