Finding solutions to improve turtle re-identification and supporting machine learning projects across Africa
Protecting the ecosystems around us is important to protecting the future of our planet and all its living citizens. Fortunately, new AI systems are conducting large-scale conservation efforts around the world, from studying animal behavior in Serengeti animals to save on declining ecosystems and preventing the discovery of poachers and wounded prey and extinction of species.
As part of our mission to help the technology we develop benefits for humanity, it is important to ensure that diverse groups of people build AI systems so that they are fair and impartial. This includes expanding the machine learning (ML) community and engaging with more audiences to address key issues using AI.
Through the research, I met Zindi, a dedicated partner with complementary goals, the biggest community of African data scientists and host competition.
The Science Team Diversity, Equity and Inclusion (DE&I) team has worked with Zindi to identify scientific challenges that will help advance conservation efforts and increase engagement in AI. Inspired by Zindi’s Bounding Box Turtle Challenge, we embarked on a project with the potential for real impact: Turtle Facial Ondation.
Biologists believe that turtles are indicator species. These are classes of behavior that help scientists understand the underlying welfare of ecosystems. For example, as the ban on chlorine pesticides in the 1970s retreated from the endangerment of extinction, the presence of otters in rivers has been considered a sign of a clean, healthy river.
The turtle is another such species. By grazing sea grass covers, they grow ecosystems and provide habitat for many fish and crustaceans. Traditionally, individual turtles have been identified and tracked by biologists with physical tags, but frequent loss or erosion of these tags in seawater has made this an unreliable method. To solve some of these challenges, we launched an ML challenge called Turtle Recall.
Given the additional challenge of keeping the turtles well enough to find tags, the Turtle Recall Challenge was intended to avoid these issues with turtle face recognition. This is possible because the turtle’s face scale patterns are unique to the individual and remain the same over decades of life.
This challenge aims to increase the reliability and speed of re-identification of turtles, and may provide a way to completely replace the use of offensive physical tags. To make this possible, we needed a dataset to work with. Luckily, after Jindy’s previous turtle-based challenge with the Kenya-based charity’s local marine conservation, the team was able to kindly share a dataset of labeled images of the turtle’s face.
The competition began in November 2021 and lasted for five months. To encourage competitors to participate, the team implemented Colab notebooks, an in-browser programming environment.
Participants were tasked with downloading challenge data and training models to predict the turtle’s identity as accurately as possible, given the photographs taken from a particular angle. We submitted predictions on data withheld from the model, allowing us to visit a public leaderboard that tracks the progress of each participant.
Community involvement was very positive, as was the technical innovations displayed by the team during the challenge. In the course of the competition, we received submissions from diverse AI enthusiasts from 13 different African countries, including countries that are not traditionally represented at the largest ML conferences such as Ghana and Benin.
Our turtle protection partners have shown that participants’ level of predictive accuracy can immediately help identify turtles on site. In other words, these models can have a true, immediate effect on wildlife conservation.
As part of Zindi’s ongoing efforts to support climate-positive challenges, they are also working on the Swahili voice classification in Kenya, helping translation and emergency services, as well as Uganda’s air quality forecasting improves social welfare.
We are grateful to Zindi’s partnership and to all those who contributed to the growth field of AI for Turtle Recall Challenge and Conservation. And we look forward to how people around the world continue to find ways to apply AI technology to build a healthy and sustainable future for the planet.
Read more about Turtle Recall on Zindi’s blog and learn more about Zindi at https://zindi.africa/
