Turtle Identification

Monitoring individual turtles is important as the more we know about where turtles are and how they are moving around, the better we can manage and conserve their populations and habitats. As a part of our research on sea turtles on Palmyra Atoll, we developed manual and automated photographic methods to identify and monitor individual animals based on the unique pattern formed by the scales on a turtle’s head.

A photographic approach provides a number of advantages as an alternative to the commonly used external flipper and internal Passive Integrated Transponder (PIT) tags as it is less costly and involves less stress to the animals. Photos taken in a known location in a turtle’s natural habitat can also contribute with additional valuable information for researchers. 

Crowdsourcing Turtle Identification

While computing power continues to rapidly increase, humans still have a huge advantage when it comes to recognizing patterns associated with details in images. To see how well people can identify individual turtles with minimal training, we tried a very simple experiment crowdsourcing the image matching effort.

Using photographs of 102 individual green sea turtles (taken between 2008 and 2013), we developed an online interface with 48,985 image pairs that volunteers reviewed to determine if it was the same animal in both images. The interface included tools for zooming, panning, and buttons for indicating whether the two images were of the same individual turtle, and registered the time it took the visitor to make their decision.

A page asking whether two photos of a turtle are depicting the same animal.

What Did We Find? 

Our visitors did pretty well! Of the image-pairs analyzed 95.69% were correctly identified. Visitors viewed an average of 33.4 image pairs (min= 1, max=1603) and viewed each image pair an average of 11 seconds before submitting their decision. The visitor with the best results viewed 176 image pairs, committed no errors, and spent an average of 1.84 seconds viewing each image pair. The next lowest error rate of 0.06% was a visitor who viewed a whopping 1,603 images, made only 1 error, and spent an average time of 6.32 seconds viewing each image pair.

These findings contribute to the development of reliable photographic approaches, where people from around the world can contribute to monitoring efforts and studies similar to ours.

Learn more about our research on Sea Turtles of Palmyra Atoll.

Acknowledgements

We would like to thank all of the anonymous visitors and the following people for their contribution to this experiment:

Margaret Regan, Hilary C., Stacy Hargrove, Hadel Go, Ariana Abrahim, Nichole Guerino, Kristen Orr, Rebecca Roberson, Elizabeth Butler, Paula Gardner, Sarah Dougher, Megan Dye, Alisha Bell, Laura Norman, Erica Fleishman, Priscilla Cole, Michelle  Trifari, Angela Mallard, Jon Becker, Case Drury, Linda DeLay, Simone Sukhdeo, Peter Houseman, Marie von Kampen, Rachel Sprague, Tritia Thamrongnawasawad, Ian McPeake, Laura López, Nate Peterson, Allison Baker, Holly Bailey, Katleen Kuhlman, Rorey Reeps, Bell  S., Iz S., Karen Nolan, Sarah Koslow, C. Ayers, Lucía Caldas,Heidi O'Neill, Sharon Bowen, Darcie Ritch, Eleanor, Sabine Glaesker, Emily Alvarez.