We invite you to participate in a mini-experiment to evaluate a test set of bottlenose dolphin (Tursiops truncatus) dorsal fin images so that we can assess the performance of different systems used for matching dorsal fin images. We are particularly interested in assessing the speed, accuracy, and usability of different photo-ID systems and hope that this experiment will provide insights into the performance of the methods and systems currently used. While we are very interested in the performance of computer vision systems for this purpose, we are also very interested in determining how the performance of these systems compares to the performance of manual matching, user-defined attribute based matching systems, and /or a combination of these or other methods. Thus, it is our hope that researchers using any type of photo-ID methods for matching images of dorsal fins participate in this experiment.
Experiment participants will be granted access to a test data set consisting of a catalog of 200 individual bottlenose dolphins, with the best cropped right or left side image of each individual, and a separate dolphin encounter (or sighting) of 50 images that contains a known number of individual animals. Experiment participants will determine the unique number of individuals within the encounter, the number of individuals in the encounter that can be matched to individuals in the catalog, and the number of 'new' individuals that should be added to the catalog. It is important that participants use whatever method or system they are currently using for dorsal-fin photo-ID and that they describe their approach in detail in the corresponding survey. For the sake of time, we have restricted this experiment to images of bottlenose dolphin dorsal fins, but plan to examine system performance with as many species as possible dependent on participant feedback.
We will circulate this experiment to several photo-ID research groups, all of which have considerable experience working in this area, as well as any interested individuals whom are actively doing photo-ID with dorsal fins. All results will remain anonymous. We will present the results of this experiment at the beginning of the “Rise of the machines - Application of automated systems for matching dolphin dorsal fins: current status and future directions” workshop to be held on Sunday, December 8, 2019, at the World Marine Mammal Conference in Barcelona, Spain, to frame the discussion topics. Results will also be posted on this website to share with the greater photo-ID community when appropriate.
If you would like to participate in this experiment please email Reny Tyson Moore (rtysonmoore@mote.org) or Kim Urian (kim.urian@gmail.com) to get access to the test data set. We anticipate it will take approximately 2 hours to complete and would appreciate that all results be submitted via this website no later than Friday, October 4, 2019, so that we have time to collate the results to present at the workshop.
Many thanks in advance for your help and participation with this endeavor,
Reny Tyson Moore & Kim Urian