ReunionTurtles Dataset

Structure and Composition

The ReunionTurtles dataset contains images of 84 individual sea turtles, comprised of 50 green sea turtles and 34 hawksbill sea turtles, all photographed during underwater expeditions. Each turtle is represented by four images, capturing both left and right profiles in two different years. This structured design enables in-depth analysis of various scientific questions, such as the differences between left and right profiles, as well as the changes in these patterns over time.

In addition to the ReunionTurtles dataset, this collection is part of a broader family of related turtle re-identification datasets, including:

• AmvrakikosTurtles: A collection featuring 50 loggerhead sea turtles, photographed during sea surveys conducted from boats.
• ZakynthosTurtles: A dataset of 40 loggerhead sea turtles, also captured underwater during ecological studies.

Scientific and Technological Applications

This dataset offers a robust foundation for designing and testing advanced algorithms in computer vision and imaging science. Specifically, the ReunionTurtles dataset can be used to explore re-identification models that detect and track individual turtles based on scale pattern recognition, even across multiple years. Researchers can delve into questions regarding the degree of similarity between opposite profiles and how scale patterns evolve over time, contributing to the development of more accurate wildlife monitoring technologies.

Complementary Resources

To facilitate the application of this dataset in re-identification models, several additional resources are available:

• wildlife-datasets: A Python library that provides a comprehensive summary of public datasets for wildlife re-identification, including two additional sea turtle datasets.
• wildlife-tools: A Python package offering tools to aid in training and evaluating re-identification models.
• Scientific Papers:
  • Paper 1: A study involving a separate dataset featuring 438 individual sea turtles.
  • Paper 2: A publication related to the Python libraries mentioned above, outlining methodologies for applying these tools to sea turtle identification.