Public dataset for archaeological site detection from Earth Observation imagery through Machine Learning
- Define a machine learning problem, come up with a project plan and a database schema
- Find locations of existing archaeological sites that are visible from public satellite basemaps, such as Google and Bing Maps
- Set up a data labeling workflow (PostGIS/QGIS)
- Perform data labeling for 100 sites
- Release an
archeoml-100dataset for the defined ML problem - Experiment, test, and decide if continue the project:
- Collaborate with other researchers to speed up labeling process
- Release new dataset versions in steps (e.g., 500, 1000, 2000)