Green Guard Challenge Summary
Our approach: In our project, we tackled the challenge of forest detection in images, contributing to the "GreenGuard" Challenge aimed at combating deforestation using AI and satellite imagery. Our approach involved leveraging time series data stored in a knowledge graph with hierarchical levels corresponding to different time stamps. Nodes in the knowledge graph were interconnected via geospatial edges, facilitating the analysis of temporal and spatial patterns. We are using a GNN to extract information about the deforestation from that knowledge graph.
Why Deforestation Matters: Deforestation contributes to climate change, biodiversity loss, and indigenous displacement. Leveraging AI and satellite imagery can enhance monitoring and prevention efforts.
Steps:
- Detect Current Forest Coverage: Use satellite imagery to map forested regions and distinguish between types of vegetation.
- Graphical representation: A knowledge graph representations of the images with the corresponding label given by our model. Each node contains several layers depending on how many time stamps of that location are available.
- Analyze Historical Data: Implement time-series analysis to detect deforestation events by identifying anomalies and changes in subsequent layers of one node and neighboring nodes.
Goal: Develop an alert and action system that not only detects potential deforestation but also provides actionable insights for conservation efforts and policymaking.
Data Sources we used:
- Eurosat Dataset for classification.
- Sentinel-1 for Science Amazonas for time series prediction and anomaly detection.
Team: Tree-mendous Troublemakers
- Julia: juliawalczak@yahoo.de
- David: david.von.nobbe@rwth-aachen.de
- Ole: ole.banning@gmail.com
- Jamison: Jamison.proctor@gmail.com
Log in or sign up for Devpost to join the conversation.