Advancing conservation with AI-based facial recognition of turtles

Finding solutions to improve identification and support automatic learning projects throughout Africa

The protection of ecological systems around us is very important to protect the future of our planet and all of its living citizens. Fortunately, the new artificial intelligence systems (AI) make progress in conservation efforts all over the world, which helps to tackle complex problems on a large scale – from studying the behavior of animal societies in Serengietti to help maintain the decrease in the ecological system, to the discovery of pipe fishing and wounded polyps to prevent species from extinction.

As part of our mission to help benefit from humanity in the technologies we develop, it is important to ensure various groups of people in future artificial intelligence systems so that they are fair and fair. This includes the expansion of the ML (ML) and participation with wider masses about addressing important problems using artificial intelligence.

Through the investigation, we encountered Zindi – a specialist partner with complementary goals – they are the largest African data scientists and host competitions that focus on solving the most urgent problems in Africa.

The work of our science team diversity team, stock diversity, and integration (De & I) with Zindi to determine the scientific challenge that can help enhance memorization efforts and participate in artificial intelligence. Inspired by the TURTLE TURTLE challenge in Zindi, we have reached a project with the possibility of obtaining a real impact: identifying the turtle’s face.

Biologists consider turtles an indicator type. These are categories of living organisms that scientists behave in understanding the basic luxury of their ecosystem. For example, the presence of foxes in the rivers was a sign of a clean healthy river, as the ban on chlorine pesticides in the seventies restored the species of the edge of extinction.

Turtle is another type of such. By grazing on seaweed cover, they plant the ecological system, and provide habitats for many fish and crustaceans. Traditionally, individual turtles have been identified and tracked by biologists with physical signs, although the frequent loss or erosion of these signs in the sea water made this unreliable method. To help solve some of these challenges, we launched a ML challenge called Turtle Recall.

Given the additional challenge of maintaining turtles is still sufficient to locate, the challenge of calling turtles aims to circumvent these problems while identifying the turtle. This is possible because the style of scales on the face of the turtle is unique to the individual and remains the same over its multi -necklace.

The challenge aims to increase reliability and the speed of identification, and may provide a way to replace the use of completely uncomfortable physical signs. To make this possible, we needed a set of data to work from it. Fortunately, after the previous challenge of the turtle -based turtle while maintaining the local charitable ocean in Kenyan, the teams preferred to share a collection of data from the pictures called turtle faces.

The competition started in November 2021 and lasted five months. To encourage the participation of competitors, the team implemented the Kulayb book, a browser programming environment, which provided a popular programming tool: Jax and Haiku.

Participants were assigned to download challenge data and training forms to predict the turtle identity, as much as possible, given a picture taken from a specific angle. After they submitted their predictions on the data that were blocked from the model, they were able to visit the general board board to track the progress of each participant.

The community’s participation was incredibly positive, as well as the artistic innovation that the teams presented during the challenge. During the competition, we received offers from a variety of fans of artificial intelligence from 13 different African countries – including countries that are not traditionally representing in the largest ML conferences, such as Ghana and Benin.

Our partners in preserving turtles indicated that the level of accuracy of the participant will be useful immediately in identifying turtles in this field, which means that these models can have a real and immediate impact on preserving wildlife.

As part of the continuous efforts of Zindi to support the positive challenges of the climate, they are also working to classify the Swahilian voice in Kenya to help translate and emergency services, and to predict air quality in Uganda to improve social care.

We are grateful to Zindi for their partnership, and everyone who contributed to their time challenges the recall of turtles and the growing field of AI to preserve. We look forward to seeing how people all over the world continue to find ways to apply artificial intelligence techniques to build a healthy and sustainable future for the planet.

Read more about calling turtles on the Zindi Blog and get to know Zindi on https://zindi.africa/