Tag a Turtle

Project description

Bonaire is a key destination for sea turtles, serving as both a nesting ground and a resting stop. This makes it an ideal location for monitoring turtle populations, which led to the creation of Sea Turtle Conservation Bonaire. For years, the organization has relied on tags and DNA tests to track turtles, but tags often fall off, and DNA testing can be costly. To address these challenges, they turned to Fontys for an innovative ICT solution. We developed an AI-powered turtle facial recognition system, enabling the team to dramatically reduce costs and save time while enhancing conservation efforts.

Context

The project centers on developing a series of Vision AI models to process images. These trained models include detection, segmentation, embedding, and environment detection. It involves extensive data acquisition and processing, model training, and the creation of AI pipelines.

Results

So far, the model has achieved an impressive 74% accuracy in identifying the correct image within the top 4 results and 84% accuracy within the top 16, all from a pool of 8,000 images. In simpler terms, what used to require the team to sift through thousands of images to find a match now only takes a few seconds, with an 84% chance that the correct image will be found within the first 16.

To make this possible, a specialized detection model was trained to recognize and crop the turtle's face with about 99% accuracy. After cropping, the image is converted into a unique numerical representation, allowing for quick and precise comparisons between images.

Thanks to this efficient process, the team can now identify individual turtles almost instantly, streamlining their work and improving the speed of their conservation efforts.

About the project group

We are a highly varied team, with most students having little to no AI experience prior to this project. The team spans across semesters 3 to 7, with members bringing backgrounds in Technology, Media, and primarily Software.