imageomics/IDLE-OO-Camera-Traps

Dataset Card for IDLE-OO Camera Traps

IDLE-OO Camera Traps is a 5-dataset benchmark of camera trap images from the Labeled Information Library of Alexandria: Biology and Conservation (LILA BC) with a total of 2,586 images for species classification. Each of the 5 benchmarks is balanced to have the same number of images for each species within it (between 310 and 1120 images), representing between 16 and 39 species.

Supported Tasks and Leaderboards

Image classification, particularly for species classification in camera trap images.

Languages

English, Latin

Dataset Structure

/dataset/
    desert-lion-balanced.csv
    ENA24-balanced.csv
    island-balanced.csv
    ohio-small-animals-balanced.csv
    orinoquia-balanced.csv
    data/test/
        desert-lion/
            <image 1>
            <image 2>
            ...
            <image 352>
        ENA24/
            <image 1>
            <image 2>
            ...
            <image 1120>
        island/
            <image 1>
            <image 2>
            ...
            <image 310>
        ohio-small-animals/
            <image 1>
            <image 2>
            ...
            <image 468>
        orinoquia/
            <image 1>
            <image 2>
            ...
            <image 336>
        metadata.csv
    notebooks/
        lilabc_CT.ipynb
        lilabc_CT.py
        lilabc_test-<dataset_name>.ipynb
        lilabc_test-filter.ipynb
        lilabc_test-filter.py
    potential-sets/
        lila-taxonomy-mapping_release.csv
        lila_image_urls_and_labels.csv
        <dataset_name>_image_urls_and_labels.csv

Data Instances

potential-sets/lila_image_urls_and_labels.csv: Reduced down to the datasets of interest listed below (from potential-sets/lila_image_urls_and_labels.csv (sha256:3fdf87ceea75f8720208a95350c3c70831a6c1c745a92bb68c7f2c3239e4c455)); all those with original_label "empty" or null scientific_name (these had non-taxa labels) were removed. Additionally, we added a multi_species column (boolean to indicate multiple species are present in the image--it gets listed once for each species in the image) and a count of how many different species are in each of those images (num_species column). This was then subdivided into CSVs for each of the target datasets (potential-sets/<dataset_name>_image_urls_and_labels.csv) in notebooks/lilabc_test-filter.ipynb. Each dataset was evaluated and sampled in its associated notebook (notebooks/lilabc_test-<dataset_name>.ipynb).

There are 184 unique scientific names in this subset (180 by full 7-rank) of those labeled at the image-level (as indicated by the CSV). This was then subdivided into CSVs for each of the target datasets (<dataset_name>-balanced.csv). These were initially identified as image-level labeled datasets and those that are a meaningful measure of our biodiversity-focused model (e.g., includes rare species--those less-commonly seen, targeting areas with greater biodiversity). The balanced datasets for each are described below.

• Desert Lion Conservation Camera Traps
  • 352 images: 32 species, with 11 images per species.
• ENA24-detection
  • 1120 images: 20 species, with 56 images per species.
• Island Conservation Camera Traps
  • 310 images: 16 species, with 10 images per species; 33 common names, 10 images per common name for all but 4 ("rooster", "petrel", "petrel chick", and "domestic chiecken"). This dataset was mostly just labeled to the family level.
• Ohio Small Animals:
  • 468 images: 39 species, with 12 images per species.
• Orinoquia Camera Traps
  • 336 images: 28 species, with 12 images per species.

Notes:

• notebooks/lilabc_CT.ipynb contains earlier analyses to understand the data provided by LILA BC (see commit fe34008).
• Not all notebooks will run under the current dataset organization (check the relative path, filenames have not changed).

Data Fields

Each of the <dataset_name>-balanced CSVs has the following columns.

• url_gcp, url_aws, url_azure are URLs to potentially access the image, we used url_aws or url_gcp.
• image_id: unique identifier for the image (provided by source).
• sequence_id: ID of the sequence to which the image belongs.
• location_id: ID of the location at which the camera was placed.
• frame_num: generally 0, 1, or 2, indicates order of image within a sequence.
• original_label: label initially assigned to the image.
• scientific_name: genus species of the animal in the image. For the island CSV, lowest rank taxa available, generally family.
• common_name: vernacular name of the animal in the image. For the island CSV, this is generally for the family, but it's a mix.
• kingdom: kingdom of the animal in the image.
• phylum: phylum of the animal in the image.
• cls: class of the animal in the image.
• order: order of the animal in the image.
• family: family of the animal in the image.
• genus: genus of the animal in the image. About half null in the island CSVs.
• species: species of the animal in the image. Mostly null in the island CSVs.
• filepath: path to the image from the data/test/ directory (<dataset-name>/<image filename>).

Notes:

• For all but the Ohio small animals dataset CSV, the images are named based on a uuid determined at the time of download. They were originally downloaded using the distributed-downloader package, so they also have the following two columns:
  • hashsum_original: MD5 hash of the original jpg image downloaded based on the CSV provided by LILA BC.
  • hashsum_resized: MD5 hash of the resized image (based on setting to resize if over 720 pixels in any dimension).
• The ohio-small-animals CSV have a filename column defined as OH_sm_animals_<filename in url_aws> and a md5 column containing the MD5 hash of the image as downloaded from the AWS bucket.
• The island-balanced CSV has an additional num_cn_images column indicating the number of images with that animal's common name.
• There is a metadata.csv included in the data/test/ directory for the dataset viewer to display images alongside their taxonomic information. The subset corresponds to the dataset-name.

Data Splits

These datasets were curated to create a small collection of camera trap image test sets.

Dataset Creation

Curation Rationale

As stated above, the goal of these datasets is to provide a collection of species classification test sets for camera trap images. Species classification within camera trap images is a real-world downstream use-case, on which a biological foundation model should be tested. These datasets were selected from those available on LILA BC since they are labeled at the image-level, and would thus not include frames labeled as containing an animal when it is simply the animal's habitat. The Island Conservation Camera Traps were of particular interest for their stated purpose of assisting in the prevention of endangered island species' extinction and the varied ecosystems represented.

Source Data

The images and their labels come from the following 5 LILA BC datasets. The labels are provided at the image level (not sequence level). Please see the source links for more information on the individual datasets.

• Desert Lion Conservation Camera Traps
• ENA24-detection
• Island Conservation Camera Traps
• Ohio Small Animals
• Orinoquia Camera Traps

Annotations

Annotations provided by the source data providers (aligned by LILA BC) are used for this test set.

Personal and Sensitive Information

These images come from an existing, public biodiversity data repository, which publishes them without associated GPS locations for the species in the images and they ensure the removal of all humans (who would otherwise have been labeled as such), so the there are no concerns.

Considerations for Using the Data

This collection of small balanced datasets was designed for testing the classification ability of BioCLIP 2 to classify species in camera trap images, a practical use-case and one on which it was not extensively trained.

Bias, Risks, and Limitations

The available species in these datasets is not a representative sample of species around the world, though they do cover a portion of species of interest to those collecting images using camera traps.

Licensing Information

This compilation is licensed under the Community Data License Agreement (permissive variant), same as the images and metadata which belong to their original sources (see citation directions below).

Citation

Please cite both this compilation and its constituent data sources:

@dataset{idle-oo-camera-traps,
  title = {{IDLE}-{OO} {C}amera {T}raps}, 
  author = {Elizabeth G Campolongo and Jianyang Gu and Net Zhang},
  year = {2025},
  url = {https://huggingface.co/datasets/imageomics/IDLE-OO-Camera-Traps},
  doi = {},
  publisher = {Hugging Face}
}

Please be sure to also cite the original data sources (provided citations on their LILA BC pages are included):

• Ohio Small Animals
  • Balasubramaniam S. Optimized Classification in Camera Trap Images: An Approach with Smart Camera Traps, Machine Learning, and Human Inference. Master’s thesis, The Ohio State University. 2024.
  • Bibtex:

    @mastersthesis{balasubramaniam2024-oh-small,
    author = {Balasubramaniam, S.},
    title = {Optimized Classification in Camera Trap Images: An Approach with Smart Camera Traps, Machine Learning, and Human Inference},
    school = {The Ohio State University},
    year = {2024},
    url = {http://rave.ohiolink.edu/etdc/view?acc_num=osu1721417695430687}
    }
    

• Desert Lion Conservation Camera Traps
  • No citation provided by source, bibtex:

    @misc{lion-ct,
    author = {Desert Lion Conservation},
    title = {Desert Lion Conservation Camera Traps},
    howpublished = {https://lila.science/datasets/desert-lion-conservation-camera-traps/},
    month = {July},
    year = {2024},
    }
    

• Orinoquia Camera Traps
  • Vélez J, McShea W, Shamon H, Castiblanco‐Camacho PJ, Tabak MA, Chalmers C, Fergus P, Fieberg J. An evaluation of platforms for processing camera‐trap data using artificial intelligence. Methods in Ecology and Evolution. 2023 Feb;14(2):459-77.
  • Bibtex:

    @article{velez2022choosing-orinoquia,
      title={Choosing an Appropriate Platform and Workflow for Processing Camera Trap Data using Artificial Intelligence},
      author={V{\'e}lez, Juliana and Castiblanco-Camacho, Paula J and Tabak, Michael A and Chalmers, Carl and Fergus, Paul and Fieberg, John},
      journal={arXiv preprint arXiv:2202.02283},
      year={2022}
    }
    

• Island Conservation Camera Traps
  • No citation provided by source, bibtex:

    @misc{island-ct,
    author = {Island Conservation},
    title = {Island Conservation Camera Traps},
    howpublished = {https://lila.science/datasets/island-conservation-camera-traps/},
    }
    

• ENA24-detection
  • Yousif H, Kays R, Zhihai H. Dynamic Programming Selection of Object Proposals for Sequence-Level Animal Species Classification in the Wild. IEEE Transactions on Circuits and Systems for Video Technology, 2019. (bibtex)
  • Bibtex:

    @article{yousif2019dynamic-ENA24,
    title={Dynamic Programming Selection of Object Proposals for Sequence-Level Animal Species Classification in the Wild},
    author={Yousif, Hayder and Kays, Roland and He, Zhihai},
    journal={IEEE Transactions on Circuits and Systems for Video Technology},
    year={2019},
    publisher={IEEE}
    }
    

Acknowledgements

This work was supported by the Imageomics Institute, which is funded by the US National Science Foundation's Harnessing the Data Revolution (HDR) program under Award #2118240 (Imageomics: A New Frontier of Biological Information Powered by Knowledge-Guided Machine Learning). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

Additionally, we would like to acknowledge and thank Labeled Information Library of Alexandria: Biology and Conservation (LILA BC) for providing a coordinated collection of camera trap images for research use.

Dataset Card Authors

Elizabeth G. Campolongo