Start with a line, let the planet complete the picture

Land Lines is an innovative project developed by Jeff Nusz and the Data Arts Team in collaboration with Zach Lieberman, designed to allow users to interact with Google Earth satellite images through gestures on their devices. The experiment enables users to "draw" or "drag" to find and connect satellite images, using Earth as a palette and their fingers as a brush. This real-time application operates in a phone's web browser without servers, leveraging machine learning, data optimization, and vantage-point trees for efficient image analysis and gesture matching. By preprocessing over fifty thousand high-resolution images with Open CV's Structured Forests edge detection and ImageJ's Ridge Detection library, the team significantly reduced the dataset to a few thousand images characterized by prominent lines. The project exemplifies big visual data exploration and connection themes, and its development insights and open-source code are accessible on GitHub.