The Present and Future of Elephant Identification
Written by: Rebecca Ebiana
In the previous article, entitled “Evolution of Elephant Identification”, I recounted how retired ELP scientist Andrea Turkalo named and came to know the forest elephants of Dzanga Bai in her 27 years of studying them. The only way we know vital demographic statistics about forest elephants is through the decades of work she did. Because understanding a population is vital to preserving it, forest elephants and the forest ecosystem of Dzanga Ndoki National Park is better off because of her work.
Before digital technology, it was more of a task to keep track of identified elephants. After digital technology, pictures were automatically saved and dated. Digital technology increased the volume of information across disciplines and worldwide, and animal behavior research is no exception. Now that this is the standard, what does it look like for scientists in Dzanga, seeking to ID elephants now?
Ana, Daniela and Julie are the intrepid scientists who are currently studying the elephants in Dzanga. Their project is to understand the animals’ infrasonic communication, or rumbles. Having the knowledge of who is making the calls is key information. But since these researchers have not spent decades in Dzanga, the elephants are not immediately familiar, and they cannot identify them on sight as Andrea did.
To aid recognition of identified elephants, Peter Wrege and Julia Gill developed a database consisting of a portion of the identified elephants photographed within the last 8 years. Ana and Julie entered and coded images with the distinguishing characteristics that Andrea used to make written identity cards–the same ear holes and rips, tusk shape, and tail features. The program is supposed to work this way: a person wanting to know who an elephant is enters a series of features and the program will return a series of likely matches. In practice, the program has similar issues that manual identification has i.e. the subjectivity and inexact nature of character identification.
Artificial Intelligence offers a solution. Matthias Körshens has created a program that uses facial recognition to identify elephants. The program uses an algorithm that takes thousands of measurements from images and learns what facial measurements are unique to an individual elephant. This would be the ideal program to keep track of identified elephants. It uses measurements which are exact, solving the problem of subjective characteristics. A user will enter a photograph and the program will return possible matches, similar to the digital database.
Although this is the ideal program, nothing is without limitations. This type of program requires a lot of data from which to learn–one must enter photographs of an elephant from different angles. The program requires a high-performance laptop to do this kind of learning, which can be expensive and impractical to use in Dzanga. However, artificial intelligence is an exciting frontier of technology that has the potential to transform science and make learning who an elephant is as simple as entering a picture into a program.