Reef-insight enables generation of detailed maps that show groups of distinct coral assemblages that can guide reef scientists in analysis and restoration projects. Moreover, our open source software framework can be extended to other regions in different parts of the world. It is an unsupervised machine learning framework via advanced clustering methods for recognition of coral reef assembles for reef community mapping via remote sensing. A wide range of clustering methods exist and it is not clear which one would be suitable for this application; hence, our framework compares major clustering approaches.
We have implemented 4 unsupervised clustering algorithms for generating benthic and geomorphic maps of One Tree island region, Great Barrier Reef, Australia.
- K-means
- Gaussian Mixture Model
- DBSCAN
- Hierarchichal Agglomerative Clustering
The repository contains two versions of the framework code:
- reef_insight_notebook.ipynb - Jupyter notebook with results
- reef_insight.py - python script implementation
S. Barve, J. Webster, and R. Chandra, "Reef-insight: A framework for reef habitat mapping with clustering methods via remote sensing", under review.