IARPA Seeks to Create Functional World Map
Intelligence research agency issues deep learning and automation challenge.
The Intelligence Advanced Research Projects Activity (IARPA), the research arm for the Office of the Director of National Intelligence, has announced the functional Map of the World (fMoW) Challenge, which officially kicks off in August. The challenge invites experts to develop deep learning and automation technologies to classify points of interest from satellite imagery. The goal is to promote research in object identification and classification to automatically identify facility, building and land use.
The challenge invites experts from across the government, academia, industry and developer communities—with or without experience in automated image analysis—to create fast and accurate classification algorithms for building and land use. The fMoW Challenge will ultimately provide a satellite imagery dataset with 1 million points of interest annotated to researchers and entrepreneurs, enabling them to better their methods for understanding satellite imagery through novel learning frameworks and multi-modal fusion techniques. Based on the data provided, participants will be asked to generate an algorithm to classify building and land use in the provided images.
Creating algorithms to accurately label parts of images has proven to be a notoriously difficult problem. The last couple of years have seen landmark advances in deep learning provide great improvements in both localization and classification of objects within scenes, according to IARPA’s written announcement. IARPA is interested to know if those techniques, or perhaps newly developed techniques, can be used to label satellite imagery accurately. A challenge provides a cost-effective way to carry out such research.
The challenge ends in December. Participants can win prizes from a total prize purse of $100,000.