GeoArgania All Trees Matters - A Geolocation Mapping Dataset of Argania Trees in the Souss Region for Building ArganiaDetector Machine Learning System
Introduction:
The GeoArgania All Trees Matters project aims to create a Geolocation Mapping Dataset of Argania trees (Argania spinosa) in the Souss region of Morocco. Leveraging Google Earth Engine, Sentinel-2 imagery, and TensorFlow Decision Forest, we intend to develop the ArganiaDetector Machine Learning System. This system will enable comprehensive analysis of the temporal and spatial dynamics of the Argania tree population from 2015 to 2022, supporting evidence-based conservation strategies and sustainable land management practices.
System Overview:
The core of the project is a comprehensive dataset containing geolocated information for 91,000 Argania trees. This dataset was meticulously collected and validated to ensure accuracy and reliability. To detect Argania trees effectively, we employ the TensorFlow Decision Forest library, specifically using a Gradient Boosted Tree model, which exhibits excellent performance with an accuracy rate of 90%.
Machine Learning Model Flow:
The machine learning model follows a structured flow to detect Argania trees based on Sentinel-2 pixel values. The steps involved in the model flow are as follows:
Application Flow:
The ArganiaDetector Machine Learning System is accessible through a user-friendly interface built using Flask. The application flow for users is as follows:
User Input: Users can upload geolocated images containing Argania trees through the interface.
ArganiaDetector System: The uploaded images are processed by the ArganiaDetector system, which triggers the machine learning model for tree detection.
Machine Learning Model: The TensorFlow Decision Forest model analyzes the Sentinel-2 pixel values in the images to identify Argania trees.
Argania Tree Detection Results: The system provides the detection results, indicating the presence and locations of Argania trees on the images.
Conclusion:
The GeoArgania All Trees Matters project endeavors to contribute to the understanding and conservation of Argania trees in the Souss region. By combining Google Earth Engine, Sentinel-2 imagery, and TensorFlow Decision Forest, we have built a powerful ArganiaDetector Machine Learning System. The comprehensive geolocation mapping dataset and high accuracy achieved by the machine learning model enable evidence-based conservation strategies and sustainable land management practices. The application of this system can foster collaboration among researchers, environmentalists, and policymakers, ensuring the protection and preservation of Argania trees, thereby highlighting the importance of "All Trees Matters" for the environment and the well-being of communities.