README.md

metadata

dataset_info:
  - config_name: default
    features:
      - name: image
        dtype: image
      - name: filename
        dtype: string
      - name: label
        dtype: string
      - name: objects
        struct:
          - name: bboxes
            sequence:
              sequence: float32
              length: 4
          - name: labels
            sequence:
              class_label:
                names:
                  '0': mosquito
    splits:
      - name: train
        num_bytes: 1753962049
        num_examples: 3138
    download_size: 1751797999
    dataset_size: 1753962049
  - config_name: raw
    features:
      - name: image
        dtype: image
      - name: filename
        dtype: string
      - name: label
        dtype: string
      - name: objects
        struct:
          - name: bboxes
            dtype:
              array2_d:
                shape:
                  - 4
                dtype: float32
          - name: labels
            sequence:
              class_label:
                names:
                  '0': mosquito
    splits:
      - name: train
        num_bytes: 1753976201
        num_examples: 3138
    download_size: 1751798435
    dataset_size: 1753976201
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
  - config_name: raw
    data_files:
      - split: train
        path: raw/train-*

This dataset is part of the Culicidaelab project - open-source system for mosquito research and analysis, which includes components:

• Data:

  • Base diversity dataset (46 species, 3139 images under CC-BY-SA-4.0 license.
  • Specialized derivatives: classification, detection, and segmentation datasets under CC-BY-SA-4.0 licenses.

• Models:

  • Top-1 models (see reports), used as default by culicidaelab library: classification (Apache 2.0), detection (AGPL-3.0), segmentation (Apache 2.0)
  • Top-5 classification models collection with accuracy >90% for 17 mosquito species.

• Protocols:

  All training parameters and metrics available at:

  • Detection model reports
  • Segmentation model reports
  • Classification experiment reports - 1st round
  • Classification experiment reports - 2nd round

• Applications:

  • Python library (AGPL-3.0) providing core ML functionality
  • Web server (AGPL-3.0) hosting API services
  • Mobile apps (AGPL-3.0): mosquitoscan for independent use with optimized models and culicidaelab-mobile for educational and research purposes as part of the CulicidaeLab Ecosystem.

These components form a cohesive ecosystem where datasets used for training models that power applications, the Python library provides core functionality to the web server, and the server exposes services consumed by the mobile application. All components are openly licensed, promoting transparency and collaboration.

This integrated approach enables comprehensive mosquito research, from data collection to analysis and visualization, supporting both scientific research and public health initiatives.

Practical Applications of the Dataset

• Scientific Research and Development:

  • Training New Models: Using the datasets to train more accurate or faster AI models tailored for specific tasks (e.g., for deployment on low-performance devices).
  • Comparative Analysis (Benchmarking): Researchers worldwide can use these datasets as a standard benchmark to compare the performance of their own detection and classification algorithms.
  • Transfer Learning: Adapting existing models to recognize mosquito species that were not included in the original dataset but are endemic to a specific region.
  • Studying Correlations: Analyzing images to identify non-obvious visual markers or relationships between species, their posture, and their environment.

• Education:

  • Educational Courses: Serving as practical material in university courses on machine learning, computer vision, and bioinformatics.
  • Training Specialists: Training future entomologists and epidemiologists to work with modern data analysis tools.

• Validation and Testing:

  • Verifying the accuracy and completeness of commercial and private insect identification systems.

License

Creative Commons Attribution Share Alike 4.0 International (CC-BY-SA-4.0)

Acknowledgments

CulicidaeLab development is supported by a grant from the Foundation for Assistance to Small Innovative Enterprises (FASIE).

Dataset Card Authors

Kovaleva Ilona

Dataset Card Contact

iloncka.ds@gmail.com