Dataset Card for Pu'u Maka'ala Phenology Transect Camera Traps (PUUM_pheno_transect_cams)

This dataset contains motion-triggered camera trap images collected at Domain 20 – Pu'u Maka'ala Natural Area Reserve, the National Ecological Observatory Network (NEON) field station in Hawai‘i. The images capture understory plant phenology alongside visits by native and invasive birds and mammals, supporting ecological monitoring, phenological analysis, and wildlife detection and classification tasks.

Dataset Details

This dataset contains 52,263 images (and counting) from motion-triggered camera traps deployed at the National Ecological Observatory Network (NEON) Domain 20 – Pu'u Maka'ala Natural Area Reserve. This is an ongoing data collection effort; currently, data have been uploaded from January 24, 2025 through October 2, 2025; however, only images from January 24 through April 4 have been processed with the detection and species ID pipeline. Images contain one or more focal individuals of NEON-monitored understory plants that can be matched to NEON data, specifically NEON data product DP1.10055.001 (Plant Phenology Observations). The data also includes various captures of native and non-native birds, as well as some nocturnal mammals.

We deployed 19 low-cost WOSPORTS G100 mini trail cameras across six NEON-monitored species (two individuals per species; 12 plants total) to capture both plant phenology and faunal activity at two vertical strata. These low-cost cameras trigger based on passive motion infrared sensor, and are primarily marketed as hunting aids for imaging animals, offering no timed capture function. Ten units were mounted on manually built tripods constructed from metal rods—adjusted to hold the lens at mid-canopy height and angled to frame the focal plant in the foreground—while nine were placed atop standard cinder-block stands at ground level. Each camera was labeled with a unique ID, configured for synchronized capture intervals, and secured in economical, weatherproof housing to withstand outdoor conditions. Each camera used two SD cards, which NEON scientists alternately swapped every 15 days during their routine transect visits. The tripod-mounted cameras recorded seasonal changes in flowers and foliage as well as arboreal visitors, whereas the cinder-block units monitored terrestrial fauna moving around the plant base.

The six species monitored and whether or not annotations for animals visiting the plants are provided are included in the below table. Note that images have only been processed through April 4, 2025.

Scientific Name	Common Name	Processed
Styphelia tameameae	pukiawe	annotations included
Vaccinium calycinum	ohelo	annotations included
Rubus hawaiiensis	akala	annotations not provided
Myrsine lessertiana	small leaf kolea	annotations not provided
Caprosma montana	pilo	annotations not provided
Cyanea Shipmaneai	haha	annotations not provided

Supported Tasks and Leaderboards

Currently an early subset of the data (from Jan 24 - April 4) has been annotated for evaluation of animal detection, across 4 cameras (ID12, ID09, ID26, ID02) representing 2 different plant taxa (pukiawe (Styphelia tameameae) and ohelo (Vaccinium calycinum)) for a total of 4,625 images (2 species of plants, with two individuals each, so 4 total plants). Using the initial combo of MegaDetectorV5 and BioCLIP 2 (run through the pybioclip package) to detect animals and filter non bird detections, we achieved a precision of .375 and recall of .76. Further work will evaluate more species detectors and classifiers as well as filtering methods, accompanying publication is in press.

Dataset Structure

This dataset is stored as a Hugging Face Dataset with images and metadata embedded in parquet files. The data can be loaded directly using the Hugging Face datasets library:

from datasets import load_dataset
dataset = load_dataset("imageomics/PUUM_pheno_transect_cams")

Repository Files

PUUM_pheno_transect_cams/
    data/
        train-00000-of-00439.parquet   # Parquet files containing images + metadata
        train-00001-of-00439.parquet
        ...
        train-00438-of-00439.parquet
    Deployment_metadata.csv                  # Camera deployment information
    PUUM_pheno_transect_cams_metadata.csv   # Standalone CSV with all metadata
    README.md

The parquet files in data/ contain both the images (as embedded bytes in the file_name column) and all associated metadata. The PUUM_pheno_transect_cams_metadata.csv provides the same metadata in CSV format for users who prefer to work with tabular data without loading the full dataset.

Original File Organization

The original images were organized by camera and collection period. This structure is preserved in the folder and filepath metadata fields:

Camera-traps/
    ID01-C/                              # Camera ID, C=cinder block, T=tripod
        R32_01_A_25_02_25_25_03_12/       # Collection period folder
            IMAG0000.jpg
            IMAG0001.jpg
            ...
        R32_01_A_25_03_26_25_04_03/
        ...
    ID02-T/
        R32_02_A_25_02_25_25_03_12/
        ...
    ...

Folder naming convention: R{SD_card_GB}_{camera_id}_{SD_card_ID}_{start_date}_{end_date}. Note that start and end dates are in YY_MM_DD format.

• SD_card_GB: Storage capacity of SD card (e.g., 32 GB)
• camera_id: Unique camera identifier
• SD_card_ID: A or B (cameras alternated between two SD cards)
• start_date / end_date: Collection period in YY_MM_DD format

Data Instances

Images are ~6000 x 4000 pixels, very noisy camera trap images. They contain a black banner along the bottom portion that contains a camera ID code and a timestamp. OCR (Optical Character Recognition) is used to automatically extract these timestamps from the same region of each image. Cameras have different cooldown periods based on the mount type: Images from tripods (mount=T) are set to capture 1 image every 5 seconds when triggered by repeated movements, while cameras on cinder blocks (mount=C) capture 3 images per trigger with a 15-second cooldown period before another trigger occurs.

Included files:

• PUUM_pheno_transect_cams_metadata.csv: Standalone CSV containing all image metadata (same fields as the parquet files, without the embedded images).

• Deployment_metadata.csv: Camera deployment-level metadata collected during field setup and maintenance.

Data Fields

PUUM_pheno_transect_cams_metadata.csv (and parquet files):

• file_name: In parquet files, contains the embedded image bytes. In the CSV, contains the original filepath.
• tags: A list of human annotations (strings) from the detection/classification pipeline at the image level. Note: Tags are only included if there is an animal detected in the image, so only 331 images from the preliminary study period (January 24 – April 4, 2025) have tags; images collected after April 4, 2025 have not been processed to include tags, meaning the lack of a tag does not mean there is no animal in those images. Tags indicate:
  • Species identifications: apapane, amakihi, omao, white-eye, kalij, unknown
  • Detection flags: no_detection_0.2 (MegaDetector confidence below 0.2), false_positives_0.2 (false positive at 0.2 threshold), missed_detection (animal present but not detected)
• timestamp: Extracted timestamp from image using OCR (pytesseract package used), may have errors.
• basename: Filename without the directory path.
• date_bin: upload period of data, or which folder it corresponds to (<start date>_<end date>, both in YY_MM_DD format).
• mount: Tripod or cinder block (T or C).
• common_name: common name of focal individual plant. There are 6 plants in total (common name (scientific name): pukiawe (Styphelia tameameae), ohelo (Vaccinium calycinum), akala (Rubus hawaiiensis), small leaf kolea (Myrsine lessertiana), pilo (Caprosma montana), and haha (Cyanea Shipmaneai)), but only 2 of them have annotations (pukiawe and ohelo).
• direction: Orientation of camera placement (W, NW, SW, E, or NE).
• camera_id: Unique identifier for the camera.
• year: Year associated to the file (2025).
• date: Full date (YYYY-MM-DD) associated with the file.
• SD_card_GB: Storage capacity (in GB) of the SD card used.
• SD_card_id: Identifier for the specific SD card used (A or B).
• valid_date: Boolean flag indicating whether the date metadata is considered valid. Mostly true if numeric or possible date.
• timestamp_check: Boolean flag for checking date falls within expected window and pertains to study period.
• folder: Name of the folder containing the file.
• filepath: Full original file path on the source filesystem.
• human_detected: Boolean flag indicating if a human was detected in the image (all false).

Deployment_metadata.csv:

• deployment-time: Timestamp indicating when the camera was initially deployed (local time: GMT-10) on January 24, 2025.
• sd-card-id: Identifier for the SD card used during the deployment period.
• camera-id: Unique identifier assigned to each camera.
• mount: Camera mounting type (tripod or cinderblock).
• common-name: Common name of the focal understory plant associated with the camera: pukiawe, ohelo, akala, small leaf kolea, pilo, or haha.
• NEON_Species_Code: NEON species code corresponding to the focal plant.
• NEON_Individual_ID_Num: NEON individual identifier used to link the plant to NEON phenology records.
• elevation: Elevation of the camera deployment location (in meters).
• elevation-units: Units associated with the elevation measurement (meters).
• distance-from-plant: Distance between the camera and the focal plant (in meters).
• distance-from-plant-units: Units associated with the distance measurement (meters).
• direction: Orientation of the camera relative to the focal plant (W, NW, SW, E, or NE).
• notes: Free-text field containing additional deployment or site-specific observations.

Data Splits

All data has been kept raw, with inference being tested but no separate splits have been generated.

Dataset Creation

Curation Rationale

This work intended to demonstrate the ability to measure phenological changes and track plant use by animals at a NEON site. See more details in accompanying paper.

Data Collection and Processing

All steps for preprocessing of data are contained in the update_guide.md in the accompanying repo. Preprocessing currently remains light, mainly consisting of OCR of timestamps and flagging of anomalous timestamps or metadata categories.

Who are the source data producers?

This data is captured at the National Ecological Observatory Network Domain 20 Pu'u Maka'ala Reserve, part of the Hawaii natural area reserves. Cameras were deployed as part of the Imageomics Institute and ABC Center Class on AI and Ecology.

Personal and Sensitive Information

This data contains endangered plants and threatened birds, so no GPS data has been included with this dataset. All images that included field technicians have been removed.

Considerations for Using the Data

Dataset Status: This is an ongoing data collection effort. The current release is a snapshot following the completion of the course by Imageomics/ABC Center titled "Experiential Introduction to AI and Ecology" in 2025. The dataset will continue to be updated with new images and annotations in the future.

Annotation Coverage: Species identification tags are only available for 331 images from the preliminary study period (January 24 – April 4, 2025). These tags were generated using MegaDetector (V5) for animal detection and BioCLIP 2 (through v2.0.0 of the pybioclip package) for species classification, followed by manual review in FiftyOne. Images collected after April 4, 2025 have not been annotated and contain empty tags as a result.

Data Collection Limitations: These images do not represent a comprehensive survey of fauna visiting the focal plants at PUUM. Missed detections, errors such as full SD cards, and camera placement all limit how much data is collected. Furthermore, the image bursts from cinder block cameras mean that the same event is potentially observed multiple times.

Bias, Risks, and Limitations

Data exhibits classic long tailed distribution. Data is noisy and captures moving plants at irregular intervals. Cameras are mounted alongside transect, an access road which we expect to be a deterrent to more shy birds and will most likely strongly impact our distributions of species. Geolocation data on deployment details metadata contains species at risk of being poached, and thus is NOT released. Annotation processing has only been completed for images through April 4, so any images from after that date do not have tags because they were not processed, not because they lacked animals. Data is preliminary and noisy, duplicates may be present.

Recommendations

Consider collection bias when attempting biological analysis or comparison with other methods.

Licensing Information

This dataset is under the Creative Commons Attribution Non Commercial 4.0 (CC-BY-NC-4.0) license.

Citation

BibTeX:

@misc{meyers2025puum,
  author = {Meyers, Luke and Potlapally, Anirudh and Long, Mike and Gabeff, Valentin and Chen, Yuyan and Jousse, Maximiliane and Zolotarev, Ted and Berger-Wolf, Tanya and Rubenstein, Daniel},
  title = {Pu'u Maka'ala Phenology Transect Camera Traps},
  year = {2025},
  url = {https://huggingface.co/datasets/imageomics/PUUM_pheno_transect_cams},
  publisher = {Hugging Face}
}

Acknowledgements

This work was supported by both the Imageomics Institute and the AI and Biodiversity Change (ABC) Global Center. The Imageomics Institute 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). The ABC Global Center is funded by the US National Science Foundation under Award No. 2330423 and the Natural Sciences and Engineering Research Council of Canada under Award No. 585136. This dataset draws on research supported by the Social Sciences and Humanities Research Council. 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, the Natural Sciences and Engineering Research Council of Canada, or the Social Sciences and Humanities Research Council.

The data were gathered at the PUUM Natural Park Reserve NEON site in Hawai‘i, in accordance with Research Permit No. 241118155000-NARS.

The National Ecological Observatory Network (NEON) is a program sponsored by the U.S. National Science Foundation and operated under cooperative agreement by Battelle. Data collected and used in this research were obtained through NEON Research Support Services.

Dataset Contributors

Luke Meyers, Anirudh Potlapally, Mike Long, Valentin Gabeff, Yuyan Chen, Maximiliane Jousse, Ted Zolotarev, Tanya Berger-Wolf, Daniel Rubenstein

Dataset Card Contact

For questions or issues, please use the Hugging Face Discussions tab or contact Luke Meyers at [email protected].