Welcome to Terrapattern, the newcomer in open source satellite imagery. We already had the American collection USGS Earth explorer and the Sentinels Scientific Data Hub of the European space agency and their research per time period. Terrapattern imagines a different system: a satellite search engine by keyword capable of detecting the same points of interest in satellite cartography (boats, swimming pools, crosswalks…). For now, in the alpha version, only the cities of New York, San Francisco, Detroit and Pittsburgh are available.
Designed within the Frank-Ratchye Studio for Creative Inquiry lab of the Carnegie Mellon university (Pittsburgh), the platform was imagined by a team headed by the artist and researcher Golan Levin, with David Newbury and Kyle McDonald. Terrapattern is relying on half a million satellite images from Open Street Map and a convolutional neural network that analyzes, characterizes images without immediately classifying them (allowing for a description of shapes, colors…).
Beyond it purely graphic appearance, the tool could have many applications and act as a citizen resource in satellite data for different actors of humanitarian aid, research, journalism as well as archeology. Terrapattern would allow you to detect sensitive points of interest (hospitals, bridges, cisterns…) to facilitate documentary and logistical operations in areas facing difficulties, like Open Street Map. The open source cartography software has become the participative tool for natural disasters (such as recently with the operation Mapping Ecuador).
Terrapattern positions itself as a scalable and participative space finding inspiration from other tools combining search programs and spontaneous contributions, such as The Signal Program of Human Security and Technology, a project from Harvard that uses satellite data to investigate war crimes and genocides, or Datakind, data mining software that has been used for instance to target microcredit aid campaigns by analyzing the proportion of straw roofs in countries of Central Africa.
For more information on Terrapattern and to access the code on Github