The amphibian portrait: individual identification of túngara frogs through image recognition software
Poster Number
7
Format
Poster Presentation
Faculty Mentor Name
Marcos Gridi-Papp
Faculty Mentor Department
Biological Sciences
Abstract/Artist Statement
Individual identification is a key aspect of research or husbandry of populations in which individuals need to be tracked across encounters. This is a challenge in many amphibians that have the body covered in soft wet skin and present no dry structures that can be tagged, clipped or stained without injuring the animal. A variety of techniques has been employed in amphibians, toe clipping being the most common. In this study we evaluated the effectiveness of photographic identification in a captive colony of túngara frogs (Engystomops pustulosus). The highly pigmented abdominal skins of túngara frogs form patterns which are distinct for each frog. We took pictures of the ventral skin in 220 adult frogs and compared these exemplar images with selections from a pool of > 800 additional pictures of the same animals using open source image-recognition software (IBEIS). We compared the similarity indices of the first and second matches in identification searches of: 1) 50 randomly selected images; 20 images of frogs with dirt (moss) attached to the skin; and 30 images of individuals fully inflated in defensive behavior. All tested individuals were identified correctly. In the random sample, the average first match had a similarity score 135 times higher than the second match. This proportion was reduced to 114 in dirty animals and increased to 177 in inflated animals. This study shows that photographic identification is robust and effective for individual identification in túngara frog populations with > 200 individuals. This method is less invasive than toe clipping, transponders or tattooing, and it can be applied to larger populations than tattooing or toe clipping. It is inexpensive, fast, does not require skilled labor, and it can be used with captive or wild populations.
Location
DeRosa University Center, Ballroom
Start Date
30-4-2016 1:30 AM
End Date
30-4-2016 3:30 PM
The amphibian portrait: individual identification of túngara frogs through image recognition software
DeRosa University Center, Ballroom
Individual identification is a key aspect of research or husbandry of populations in which individuals need to be tracked across encounters. This is a challenge in many amphibians that have the body covered in soft wet skin and present no dry structures that can be tagged, clipped or stained without injuring the animal. A variety of techniques has been employed in amphibians, toe clipping being the most common. In this study we evaluated the effectiveness of photographic identification in a captive colony of túngara frogs (Engystomops pustulosus). The highly pigmented abdominal skins of túngara frogs form patterns which are distinct for each frog. We took pictures of the ventral skin in 220 adult frogs and compared these exemplar images with selections from a pool of > 800 additional pictures of the same animals using open source image-recognition software (IBEIS). We compared the similarity indices of the first and second matches in identification searches of: 1) 50 randomly selected images; 20 images of frogs with dirt (moss) attached to the skin; and 30 images of individuals fully inflated in defensive behavior. All tested individuals were identified correctly. In the random sample, the average first match had a similarity score 135 times higher than the second match. This proportion was reduced to 114 in dirty animals and increased to 177 in inflated animals. This study shows that photographic identification is robust and effective for individual identification in túngara frog populations with > 200 individuals. This method is less invasive than toe clipping, transponders or tattooing, and it can be applied to larger populations than tattooing or toe clipping. It is inexpensive, fast, does not require skilled labor, and it can be used with captive or wild populations.