Work Description

Title: Skelevision Dataset Open Access Deposited

O
Attribute Value
Methodology
  • Methodology: Bird skeletal specimens were photographed from a consistent perspective with a semi-haphazard layout. All photographs were taken from 400 mm above the surface; an RGB image (5,472 x 3,468 pixel resolution) was captured with a FLIR Blackfly S camera (with a SONY IX183 sensor) and a 3D image was captured with an Intel RealSense D415 active stereoscopic camera (a 1,280x720 pixel active depth resolution and a 1,920 x 1,080 RGB resolution). Prior to photographing, the keel and skull were consistently oriented so that a profile view was captured in the image, and the specimen ID tag was situated to ensure that the catalog ID number was visible. The remainder of the bones were then haphazardly spread across the surface using a paintbrush to minimize overlap of the bones, but without any particular orientation.
Description
  • Description: Each folder contains all of the data for a specific specimen; the folder names correspond to the University of Michigan Museum of Zoology catalog number for the specimen. Folders with a “-“ in the name are individual specimens that were photographed multiple independent times; the number following the “-“ indicates the repetition number (i.e. the folder named “UMMZ_242382-10” contains the tenth set of photographs for specimen UMMZ 242382). The photographs are necessary to train and test the Skelevision model, which is a computer vision approach to identifying and measuring elements of the skeleton (length of the tibiotarsus, tarsometatarsus, femur, humerus, ulna, radius, carpometacarpus, 2nd digit 1st phalanx, skull, and keel; the outer diameter of the sclerotic ring at its widest point; and the distance from the back of the skull to the tip of the bill). The data span 115 species of passerines across 79 genera from 59 families.
Creator
Depositor
  • bwbenz@umich.edu
Contact information
Discipline
Funding agency
  • Other Funding Agency
Other Funding agency
  • UM
Keyword
Date coverage
  • 2020 to 2022
Citations to related material
  • Weeks, B.C., Zhou, Z., O’Brien, B., Darling, R., Dean, M., Dias, T., Hassena, G., Zhang, M., and Fouhey, D.F. 2022. A deep neural network for high throughput measurement of functional traits on museum skeletal specimens. Methods in Ecology and Evolution. https://doi.org/10.1111/2041-210X.13864
Resource type
Last modified
  • 01/29/2023
Published
  • 03/18/2022
Language
DOI
  • https://doi.org/10.7302/dtkh-qw53
License
  • UMMZ Bird Division Digital Data Usage Agreement (Draft) - attached below
To Cite this Work:
Brian C. Weeks. (2022). Skelevision Dataset [Data set], University of Michigan - Deep Blue Data. https://doi.org/10.7302/dtkh-qw53

Relationships

In Collection:

Files (Count: 4; Size: 35.7 GB)

Thumbnailthumbnail-column Title Original Upload Last Modified File Size Access Actions
Specimens_Examined.csv 2022-03-15 2022-03-15 57.6 KB Open Access
UMMZ_Bird_Division_Digital_Data_...t.pdf 2022-03-15 2022-03-15 73.4 KB Open Access
readme.txt 2022-03-16 2022-03-16 2.23 KB Open Access
UMMZ_Bird_Skeleton_Images.zip 2022-03-17 2022-03-18 35.7 GB Open Access

High Resolution RGB Photographs
Within each folder there is a .tiff file. These files have been saved with the UMMZ catalog number in the filename (skeleton-UMMZCATALOG#-Color2.tiff). The files are high resolution (5,472 x 3,468 pixels) images of the specimen captured with the FLIR Blackfly S camera (with a SONY IX183 sensor). These are the images that were used to train and validate the Skelevision model.

For each image, the specimen was randomly placed on the surface, the skull and keel were oriented so that their profiles were captured in the image, the associated specimen tag was placed so that the catalog number was visible, and the remaining bones were spread haphazardly across the surface using a paintbrush.

Stereoscopic Depth Camera RGB File
Within each folder there is a Color.png file. These files have been saved with the UMMZ catalog number in the filename (skeleton-UMMZCATALOG#-Color.png). The files are the RGB images (1,280 x 720 pixels) captured by the Intel RealSense stereoscopic camera. These files were not used in the Skelevision model, but have been retained for future use.

Stereoscopic Camera Depth Image
Within each folder there is a Depth.png file. These files have been saved with the UMMZ catalog number in the filename (skeleton-UMMZCATALOGX-Depth.png). The files are an RGB image of distance-to-surface data generated by the Intel RealSense camera. These files were not used in the Skelevision model, but have been retained for future use as-needed.

Stereoscopic Depth Camera Data
Within each folder there is a.csv file. These files have been saved with the UMMZ catalog number in the filename (skeleton-UMMZCATALOG#-Depth-data.csv). The files contain depth data representing the distance from the camera to the surface for each pixel in the scene captured by the Intel RealSense Camera (corresponding -Depth.png and -Color.png files). These data are not absolute values of distance, but can be converted to 3D data, in combination with the RGB image files generated for each specimen. These files were not used in the Skelevision model, but have been retained for future use as-needed.

To Cite this Work:
Brian C. Weeks. Skelevision Dataset [Data set], University of Michigan - Deep Blue Data. https://doi.org/10.7302/dtkh-qw53

Download All Files (To download individual files, select them in the “Files” panel above)

Total work file size of 35.7 GB is too large to download directly. Consider using Globus (see below).

Files are ready   Download Data from Globus
Best for data sets > 3 GB. Globus is the platform Deep Blue Data uses to make large data sets available.   More about Globus

Remediation of Harmful Language

The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.