Hugging Face's logo

Datasets:
imageomics
/
2018-NEON-beetles

Tasks:
Image Classification
Image Segmentation
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
1K - 10K
Tags:
biology
image
animals
CV
beetles
elytra
Libraries:
Datasets
License:
Dataset card Viewer Files Community
17
2018-NEON-beetles / README.md
egrace479's picture
egrace479
change configs for resize as suggested in #8
dcb9e1e verified
preview code
|
raw
history blame contribute delete
25.2 kB
metadata
license: cc-by-sa-4.0
language:
  - en
pretty_name: 2018 NEON Ethanol-preserved Ground Beetles
task_categories:
  - image-classification
  - image-segmentation
tags:
  - biology
  - image
  - animals
  - CV
  - beetles
  - elytra
  - elytra width
  - elytra length
  - size
size_categories:
  - n<1K
  - 10K<n<100K
configs:
  - config_name: group_images
    data_files:
      - BeetleMeasurements.csv
      - group_images/*.jpg
  - config_name: resized_images
    data_files:
      - beetle_images_resized/*.jpg

Dataset Card for 2018 NEON Ethanol-preserved Ground Beetles

Collection of ethanol-preserved ground beetles (family Carabidae) collected from various NEON sites in 2018 and photographed in batches in 2022.

Dataset Details

Dataset Description

This dataset is composed of a collection of 577 images of ethanol-preserved beetles collected at NEON sites in 2018. Each image contains a collection of beetles of the same species from a single plot at the labeled site. In 2022, they were arranged on a lattice and photographed; the elytra length and width were then annotated for each individual in each image using Zooniverse.

The individual images were segemented out based on scaling the elytra measurement pixel coordinates to the full-size images (more information on this process is available on the Imageomics/2018-NEON-beetles-processing repository).

Dataset Structure

In group_images we have 577 images of multiple beetles in a reasonably regular pattern (see sample image below). Each of these images contains a centimeter-based scalebar and a barcode representing the sample (collection of beetles in the image). Each row in BeetleMeasurements.csv corresponds to a measurement of the elytra (length or width) of an individual beetle.

The individual_images directory contains images of each individual segmented from the group images. These connect back to their group image based on the individualID.

/dataset/
    group_images/
        <pictureID 1>.jpg
        <pictureID 2>.jpg
        ...
        <pictureID 577>.jpg
    beetle_images_resized/
        <pictureID 1>.jpg
        <pictureID 2>.jpg
        ...
        <pictureID 577>.jpg
    individual_images/
        <individualID 1>.jpg
        <individualID 2>.jpg
        ...
        <individualID 11459>.jpg
    BeetleMeasurements.csv
    BeetleMeasurements_resized.csv
    individual_metadata.csv

Data Instances

[More Information Needed]

The group_images folder has all 577 orignal full size images. The beetle_images_resized folder has the resized (approximately 1MB) images used for annotation in Zooniverse (1MB was the size limitation imposed by the software); aspect ratio was maintained in the resize.

Segmentation was completed by adapting the elytra length and width annotations to generate the xy point input for the Segment Anything (SAM) model. Segment Anything (SAM) was then used to segment the individual beetles. More information and the code for this process is provided in the Imageomics/2018-NEON-beetles-processing repo.

picture of beetles in A00000046094
Image of Carabus goryi beetles from sample A00000046094.

Data Fields

BeetleMeasurements.csv:

  • pictureID: Name of the image (<sample-barcode>.jpg, though sometimes it's <NEON_sampleID_>.jpg); this matches the barcode on the sample bottle (included in the image). Unique identifier for the 577 group images, not for dataset. There are 11,459 individual beetles. All individuals in a single image are from the same location and are the same species. However, there is one image (MOAB_001.S.20180724.jpg) with 17 individuals that does not have any taxonomic information. 10 individuals do not have species-level labels
  • scalebar: Pixel coordinates indicating the beginning and end of one side of a 1cm square on the ruler/scalebar in the image. This is indicated by two points p1 = (x1, y1) and p2 = (x2, y2), such that the Euclidean distance between them is the number of pixels in one centimeter.
  • cm_pix: Integer. The number of pixels in a centimeter on the scalebar (measured by Euclidean distance between points given in scalebar).
  • individual: Integer. The beetle in the image to whom the measurements refer. The beetles are measured "in line" (numbering is from left to right, top to bottom, as one would read a book). However, Zooniverse does not register individual IDs beyond 99 (once 99 individuals are marked, it restarts the count at 1). Be careful with it.
  • structure: Whether the measurement applies to the length or width of the elytra (ElytraLength or ElytraWidth, respectively).
  • lying_flat: Whether or not the beetle is lying flat in the image (Yes or No). This does NOT correlate to vertical or horizontal alignment; it is whether the individual is twisted to one side or not--a side effect of how they died. Generally, the measured width will be shorter than the true width if this is No.
  • coords_pix: Pixel coordinates of the line marking the length or width of the elytra (green or purple line in the sample image). Note that these annotations were done on the resized copy of the image (1MB), but the aspect ratio was maintained in the resizing.
  • dist_pix: Float. The length or width of the elytra (indicated by structure) as measured in pixels (calculated as Euclidean distance beetween coords_pix points).
  • dist_cm: Float. The length or width of the elytra (indicated by structure) as measured in centimeters using the scalebar compared to the measured pixel distance (dist_pix/cm_pix).
  • scientificName: Scientific name of the specimen (<Genus> <species>). There are 78 different species among 36 genera, though another 10 individuals are only labeled to the genus (or subgenus) level. The 17 individuals in PictureID MOAB_001.S.20180724.jpg are missing the scientific name label (they are all the same species). These labels were provided by NEON.
  • NEON_sampleID: NEON identifier for the sample (576 unique IDs), prefixed by the plotID.
  • siteID: String. Identifier for the site from which the specimens were collected. There 30 different site IDs. For more information on the sites, visit: https://www.neonscience.org/field-sites/explore-field-sites.
  • site_name: Name of field site from which the specimens were collected. There are 43 different field site names. -- These lack consistency (sometimes "NEON" or "Site" was used in the name causing duplicates), and will be updated using the site codes.
  • plotID: Identifier for the plot from which the specimens were collected (<siteID>_<plot number>). Each site has multiple plots; there are 144 total plots from which specimens were collected.
  • user_name: Zooniverse username of person annotating the elytra length and width information (there were 3 annotators).
  • workflowID: Integer identifier for the Zooniverse workflow used. There are 5.
  • genus: Genus of the individual (generated by taking the first word in the scientificName). There are 36 unique genera labeled.
  • species: Species of the individual (generated from the word(s) following the genus in the scientificName). There are 78 unique species labeled.
  • combinedID: Generated from PictureID (minus the .jpg) plus _<individual>. This gives only 11,104 IDs, as it suffers from the issue of individual based on Zooniverse's export. Individuals should be measured by half the number of rows with user_name == "IsaFluck".
  • measureID: Unique identifier for the beetle measurements (and this CSV), generated using the uuid package. There are 39,064 unique measurements (2 per beetle, with overlap on 234 images annotated by all three annotators and 1 annotated by just two).
  • file_name: Relative path to image from the root of the directory (<group_images>/<pictureID>.jpg); allows for image to be displayed in the dataset viewer alongside its associated metadata.
  • image_dim: Dimensions of full-sized image (<height>, <width>, <number of channels>).
  • resized_image_dim: Dimensions of resized image (<height>, <width>, <number of channels>).
  • coords_pix_scaled_up: coords_pix adjusted for full-sized image.

NOTE:

  • The NEON_sampleID RMNP_014.20180709.CALADV.01 is repeated because there were too many individuals in the sample to organize them all in one picture. Thus, the individuals from this sample are split between two pictures: A00000051555_1 and A00000051555_2.
  • The NEON_sampleID MOAB_001.S.20180724 was provided without scientific name identification. This sample is Cicindela punctulata punctulata (confirmed in NEON biorepository records by Laura Nagel).
  • See notebooks in the Beetle Processing Repository for more information and analyses of these values, addition and renaming of the above columns from the original measurement CSV, and the creation of the individual metadata CSV.
  • BeetleMeasurements_resized.csv has pictureID, scientificName, genus, species, NEON_sampleID, siteID, and file_name columns to display the resized images in the dataset viewer with some group image-level information. It will match up to information from BeetleMeasurments.csv on pictureID, though this is not a unique alignment since it's not a unique ID for BeetleMeasurements.csv, but is for the resize.

individual_metadata.csv:

  • individualID: ID of beetle in the individual image (<measureID_length>_<measureID_width>). This is a unique identifier for this CSV.
  • combinedID: Generated from PictureID (minus the .jpg) plus _<individual>. (Matches combinedID in BeetleMeasurements.csv.)
  • lying_flat: Whether or not the beetle is lying flat in the image (Yes or No). This does NOT correlate to vertical or horizontal alignment; it is whether the individual is twisted to one side or not--a side effect of how they died. Generally, the measured width will be shorter than the true width if this is No.
  • elytraLength: Length of the elytra in centimeters. Indicated by the green line in the image below.
  • elytraWidth: Width of the elytra in centimeters. Indicated by the purple line in the image below.
  • measureID_length: measureID from BeetleMeasurements.csv for the elytraLength of this individual. Can be used to fetch measure-specific information from BeetleMeasurements.csv.
  • measureID_width: measureID from BeetleMeasurements.csv for the elytraWidth of this individual. Can be used to fetch measure-specific information from BeetleMeasurements.csv.
  • genus: Genus of the individual (generated by taking the first word in the scientificName from BeetleMeasurements.csv). There are 36 unique genera labeled.
  • species: Species of the individual (generated from the word(s) following the genus in the scientificName from BeetleMeasurements.csv). There are 78 unique species labeled.
  • NEON_sampleID: NEON identifier for the sample (576 unique IDs), prefixed by the plotID. (Matches NEON_sampleID in BeetleMeasurements.csv.)
  • file_name: Relative path to image from the root of the directory (individual_images/<individualID>.jpg); allows for image to be displayed in the dataset viewer alongside its associated metadata.
Figure 1
Figure 1. Sample image of beetles with green and purple lines to indicate the measurement used for the elytra length and width, respectively. The red line marks the pixels for 1cm.

Data Splits

This dataset does not have train-test splits.

Dataset Creation

Curation Rationale

[More Information Needed]

This dataset was curated for use at BeetlePalooza 2024 to help drive scientific discovery through machine learning and computer vision.

Source Data

The specimens come from NEON sites, as indicated in the BeetleMeasurements CSV. They were identified by domain scientists (beetle experts) based on physical examination. Each sample consists of a single species. For more information about general NEON data, please see their Ground beetles sampled from pitfall traps page.

The samples were sent from NEON, preserved in ethanol, then arranged on a lattice to be photographed. Isadora Fluck photographed the beetles in each sample in 2022, using a Nikon D500 with lens: AF-S Micro NIKKOR 60mm 1:2.8G ED (Nikon).

Data Collection and Processing

Beetles were collected in 2018 by NEON field technicians. In sum, Ground beetles are sampled using pitfall traps (16 oz deli containers filled with 150 or 250 mL of propylene glycol). Multiple traps are deployed in each of 10 plots at each terrestrial NEON site (from 2013-2017, 40 traps per site; 2018-present, 30 traps per site), with traps arrayed approximately 20 meters from the center of the plot in one of the four cardinal directions. Sampling occurs biweekly throughout the growing season (when temperatures are above 4 degrees C). The corresponding NEON data product for these data is entitled, "Ground beetles sampled from pitfall traps". More complete field and biorepository protocols can be found at this link.

This dataset specifically focuses on 2018 because at the time the images were taken in 2022, 2018 had the most complete spatial coverage of identified and fully processed beetle specimens (due to latency issues lingering from the pandemic and the time taken for experts to ID difficult species). Images were only taken of specimens stored in ethanol within vials as it was too difficult to ship the pinned specimens to the University of Florida where imaging took place.

The full citation for the NEON ground beetle data for the closest release year to the time of sampling is: NEON (National Ecological Observatory Network). Ground beetles sampled from pitfall traps (DP1.10022.001), RELEASE-2021. https://doi.org/10.48443/tx5f-dy17. Dataset accessed from https://data.neonscience.org/data-products/DP1.10022.001/RELEASE-2021 on August 11, 2024.

Who are the source data producers?

This dataset is a collection of images taken from beetles collected by NEON field technicians that were preserved in ethanol following collection from NEON sites in 2018. A subset of difficult to ID taxa were pinned and sent to experts for identification. The images provided here were of more common species that did not require expert ID that could be identified by NEON field technicians. After being preserved in ethanol, the beetles were archived at the [NEON Biorepository] (https://biorepo.neonscience.org/portal/). They were were sent by Nico Franz, Kelsey Yule, and Andrew Johnston from the NEON Biorepository to Ben Baiser, Sydne Record, and Isadora Fluck for imaging in 2022. Information regarding the sites from which they were collected and the taxonomic labels were provided by NEON. Site information (names and ecoclimatic domains) can be matched to siteID from the NEON_Field_Site_Metadata_20240802.csv, which is available on NEON's field sites information page (click Download Field Site Table (CSV)).

Annotations

Annotations (elytra length and width) were completed in Zooniverse by Isadora Fluck and two Harvard Forest summer REU students (Riley Wolcheski and Isha Chinniah); annotator is indicated by their NEON system usernames in the BeetleMeasurements CSV. The annotations (2 per beetle) were repeated for a subset of the images to measure observation error introduced in the annotation process. The taxonomic labels were provided by NEON with the samples.

Annotation process

The images were downsized to 1MB (preserving aspect ratio) for annotation in Zooniverse. Overall, 234 images were annotated by all three annotators and 1 was annotated by just two. Beetles were indicated as not lying flat if they tended to one or the other side, as this can result in a smaller width measurement than expected. See the sample image (also in Figure 1) for indicators of length and width, as well as the centimeter mark in pixels on the scalebar.

Who are the annotators?

  • Isadora E. Fluck: Ph.D. Candidate at Baiser Lab of Community Ecology University of Florida annotated all samples, and led annotation effort of two summer REU students under the supervision of Sydne Record:
    • Riley Wolcheski (University of Connecticut)
    • Isha Chinniah (Mount Holyoke College)

Personal and Sensitive Information

N/A

Considerations for Using the Data

  • The NEON_sampleID RMNP_014.20180709.CALADV.01 is repeated because there were too many individuals in the sample to organize them all in one picture. Thus, the individuals from this sample are split between two pictures: A00000051555_1 and A00000051555_2.
  • The NEON_sampleID MOAB_001.S.20180724 was provided without scientific name identification.
  • The individual indicator is not unique to pictureID since Zooniverse restarted IDs after individual number 99, so individuals are indicated by measurements annotated by user_name == "IsaFluck" since she annotated each image once, with the other annotators just labeling a subset for comparison.
  • These images do not include all beetles sampled at all sites in 2018 because they do not include the pinned specimens.
  • Dorsal images may not provide all the information needed to ID a beetle.

Bias, Risks, and Limitations

This dataset does not have a balanced representation of genera. In particular, rare or more difficult to ID taxa are not included here. These images do not include all beetles sampled at all sites in 2018 because they do not include the pinned specimens.

Recommendations

Be aware of the distribution when determining use cases and metrics for evaluation.

Licensing Information

This dataset is licensed under CC BY-SA.

Citation

BibTeX:

Data

@misc{Fluck2018_NEON_Beetle,
  author = {Isadora E. Fluck and Benjamin Baiser and Riley Wolcheski and Isha Chinniah and Sydne Record},
  title = {2018 NEON Ethanol-preserved Ground Beetles},
  year = {2024},
  url = {https://huggingface.co/datasets/imageomics/2018-NEON-beetles},
  doi = {<doi once generated>},
  publisher = {Hugging Face}
}

If you use the individual segmented images, please also cite the code repository (Imageomics/2018-NEON-beetles-processing) used to produce them.

Acknowledgements

The National Ecological Observatory Network is a program sponsored by the National Science Foundation and operated under cooperative agreement by Battelle. This material uses specimens and/or samples collected as part of the NEON Program. Additionally, this material is based upon work supported by the National Science Foundation under Award Numbers 2301322, 1950364, and 1926569, as well as the USDA National Institute of Food and Agriculture, Hatch Project Number ME0-22425 through the Maine Agricultural and Forest Experiment Station. Additional support was provided by the Imageomics Institute, which is supported by NSF Award No. 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.

Dataset Card Authors

Elizabeth G. Campolongo