---
task_categories:
- image-classification
tags:
- butterfly
- separated wings
- dorsal
- ventral
- RGB
- CV
pretty_name: Smithsonian Tropical Research Institute (STRI) Samples
size_categories:
- n<1K
configs:
- config_name: default
  data_files:
  - metadata.csv
  - images/*.JPG
license: cc-by-4.0
---

# Dataset Card for Smithsonian Tropical Research Institute (STRI) Samples

## Dataset Description

- **Curated by:** Christopher Lawrence, Owen McMillan, Daniel Romero, and Carlos Arias
<!-- Provide the basic links for the dataset. These will show up on the sidebar to the right of your dataset card ("Curated by" too). -->
<!-- - **Homepage:** -->
- **Repository:** https://github.com/Imageomics/wing-segmentation
<!-- - **Paper:** -->

### Dataset Summary

Dorsal images of butterfly wings collected by [Owen McMillan](https://stri.si.edu/scientist/owen-mcmillan) and members of his lab at the Smithsonian Tropical Research Institute.

Full dataset will be 24,119 RGB images: Dorsal and Ventral images of separated wings. This sample contains 207 _dorsal_ butterfly images used as part of the training data for [Imageomics/butterfly_detection_yolo](https://huggingface.co/imageomics/butterfly_detection_yolo).
<!---
This dataset card was generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/datasetcard_template.md?plain=1), and further altered to suit Imageomics Institute needs.
--->


### Supported Tasks and Leaderboards

Note that this is just a subset of a larger dataset that has yet to be curated. It was used in combination with images from other sources to train the [Imageomics/butterfly_detection_yolo](https://huggingface.co/imageomics/butterfly_detection_yolo) model, and is not large or representative enough for use in training on its own yet.

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->


```
/dataset/
    <images>/
        <img_name_no_ext 1>.JPG
        <img_name_no_ext 2>.JPG
        ...
        <img_name_no_ext 207>.JPG
    metadata.csv
```


### Data Instances

<!--
Describe data files

Ex: All images are named <img_id>.png, each within a folder named for the species. They are 1024 x 1024, and the color has been standardized using <link to color standardization package>.
-->

* **Type:** JPG
* **Background (color or none):** Sometimes gray sometimes white
* **Fit in frame:** tight
* **Ruler or Scale:** No
* **Color:** 5x6 Color checker

Potential for future addition of genetic sequencing data.

There are 11 species represented in this sample of 207. They are primarily from the genus Heliconius, but the following genera are also represented: Junonia, Eueides, Neruda, and Dryas.

The CSV also includes labels regarding the quality of the specimen: `damaged`, `dorsal`, or `incomplete`, which were used for the segmentation model. Below is a species distribution chart for this data colored by quality label.

![species distribution chart for STRI data colored by label](https://github.com/user-attachments/assets/92da2996-b4f0-462a-b666-4ea715669f3b)

### Data Fields

**metadata.csv:** minimal data record for the images; this is a subset of all the information potentially available for each image since curation of this dataset is still ongoing. This also means that there are some duplicative columns (e.g., there are multiple NCSU and STRI IDs). 
- **label:** indicates quality of specimen, either `damaged`, `dorsal`, or `incomplete`.
- **img_name_no_ext:** `FileName` without the type extension.
- **source_url:** not actually a URL, but a box location, with the folder.
- **source_dataset:** All `STRI`, holdover from combination into larger dataset.
- **FileName:** filename given to image (with extension, e.g., ".JPG"). Often `<NCSU_id>_D.JPG`.
- **Folder:** folder in which images or a record is kept, seems to be date followed by number; part of the "source_url".
- **Box:** Box in which specimen was kept ("Box 11" or "Box 10").
- **Genus:** genus of the specimen; primarily "Heliconius", but also "Junonia", "Eueides", "Neruda", or "Dryas". Labeled for all images.
- **species:** species of the specimen; 11 unique species, but mostly "erato" and "melpomene". Labeled for all but six of the images.
- **subspecies:** subspecies of the specimen if provided, or parent subspecies if hybrid. Labeled for 168 images.
- **subspecies II:** second parent subspecies of the specimen if it is a hybrid (there are 16 with the label "hybrid"). Labeled for 70 images.
- **Subfamily**: subfamily of the specimen (only labeled for 18 images, all "Heliconiinae").
- **id uncertain:** indicates uncertainty in the taxonomic identity of the specimen? -- 11 images, but there is another image with a question mark in the subspecies name.
- **sex:** sex of the specimen ("male" or "female", one instance of "?", most labeled).

**Columns maintained for future clean-up and recombination**
- `MULTIPLE IDS` -- keep for alignment and standardization later
- `NCSU_id`, ID from NCSU? Did all values have this or is it just NCSU prefix?
- `NCSU_num` (previously `NCSU #`) --relation to NCSU ID and STRI ID?
- `stri_id`, --relation to NCSU id?
- `STRI ID`, -- relation to other versions above?
- `Other ID`,-- relation to other versions above?


- **Date Collected:** date specimen was collected? Seem to be mostly DD-MM-YY, but not consistent.
- **Collector:** person who collected the specimen?
- **Taxonomic ID:** --- only labeled for 18 images, what is this? Seems to be scientific name: "Genus species (ssp x ssp II)" or "Genus species (hybrid)". Remove or do same for the remaining images?
- **Sample type:** "Wild" or "Stocks", only labeled for 18 images.
- **Owner of Sample:** -- only labeled for 18 images, mostly "BLUM - STRI", but also two "MCMILLAN - STRI", corresponds to Collector for BLUM, but not for MCMILLAN.


### Data Splits

All of these sample images were used in training the [Imageomics/butterfly_detection_yolo](https://huggingface.co/imageomics/butterfly_detection_yolo) model.

## Dataset Creation

### Curation Rationale

These images were selected and labeled to use in training the [Imageomics/butterfly_detection_yolo](https://huggingface.co/imageomics/butterfly_detection_yolo) model. The larger dataset is still undergoing curation.

### Source Data

Images are of specimens collected by the [Smithsonian Tropical Research Institute](https://stri.si.edu/) and photographs of this subset were taken by Daniel Romero, Carlos Arias, and other members of the Owen McMillan Lab at the Smithsonian Tropical Research Institute Gamboa Station.
<!-- images taken by Christopher Lawrence have not been added yet, curation ongoing -->

#### Data Collection and Processing
[More Information Needed]

Photographs of the wings were taken using an Olympus OM-D EM-1 digital camera with an Olympus Zuiko Digital ED 60 mm f/2.8 macro lens.

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, re-sizing of images, tools and libraries used, etc. 
This is what _you_ did to it following collection from the original source; it will be overall processing if you collected the data initially.
-->

#### Who are the source data producers?
<!-- This section describes the people or systems who originally created the data.

Ex: This dataset is a collection of images taken of the butterfly collection housed at the Ohio State University Museum of Biological Diversity. The associated labels and metadata are the information provided with the collection from biologists that study butterflies and supplied the specimens to the museum.
 -->
The dataset is a collection of images taken of the McMillan Lab's collection of butterfly samples obtained from field collection surveys and several raised in captivity for genetic and other studies.

### Annotations
There are several annotations in this data that are not standard and are very specific. Further annotation is needed. In general, taxonomic information down to the species and subspecies name is provided but is not always correct and needs double checking at this moment.

#### Annotation process
Damage annotations were completed by Michelle Ramirez for training the [Imageomics/butterfly_detection_yolo](https://huggingface.co/imageomics/butterfly_detection_yolo) model she created.

#### Who are the annotators?

<!-- This section describes the people or systems who created the annotations. -->
Annotation data (other than the damage labels) was provided by the McMillan Lab.

### Personal and Sensitive Information
[More Information Needed]
<!-- 
For instance, if your data includes people or endangered species. -->


## Considerations for Using the Data
<!--
Things to consider while working with the dataset. For instance, maybe there are hybrids and they are labeled in the `hybrid_stat` column, so to get a subset without hybrids, subset to all instances in the metadata file such that `hybrid_stat` is _not_ "hybrid".
-->

### Bias, Risks, and Limitations
This is a small sample of minimally curated butterfly images intended only for wing segmentation and damage detection. It is **not** intended for species classification.

### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

This is not a complete dataset, nor is it curated for greater use. The purpose of this dataset is to provide the 207 image sample used as part of a larger training set for the [Imageomics/butterfly_detection_yolo](https://huggingface.co/imageomics/butterfly_detection_yolo) model. More information on the full project can be found in the [GitHub repository](https://github.com/Imageomics/wing-segmentation).

Recommended use is only for segmentation.


## Additional Information

### Dataset Curators

* Christopher Lawrence (Princeton University) - ORCID: 0000-0002-3846-5968
* Owen McMillan (Smithsonian Tropical Research Institute) - ORCID: 0000-0003-2805-2745
* Daniel Romero (Smithsonian Tropical Research Institute)
* Carlos Arias (Smithsonian Tropical Research Institute) - ORCID: 0000-0002-1546-5156

### Licensing Information

[CC BY](https://creativecommons.org/licenses/by/4.0/)

### Citation Information

Christopher Lawrence, Owen McMillan, Daniel Romero, Carlos Arias. (2024). Smithsonian Tropical Research Institute (STRI) Samples. Hugging Face. https://huggingface.co/datasets/imageomics/STRI-Samples.

## Acknowledgements

This work was supported by US National Science Foundation (NSF) grant [IOS 2110532](https://www.nsf.gov/awardsearch/showAward?AWD_ID=2110532&HistoricalAwards=false), with additional support from the [Imageomics Institute](https://imageomics.org), which is funded by the NSF's Harnessing the Data Revolution (HDR) program under [Award #2118240](https://www.nsf.gov/awardsearch/showAward?AWD_ID=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.