Deep learning for label-scarce remote sensing applications

Abstract:

State-of-the-art deep learning methods require large quantities of labeled data pairs for high performance. While satellite data is now available in abundance, ground truth labels remain scarce. Moreover, ground truth labels are distributed unevenly around the globe; high-resource regions (e.g. US, Europe) have many more labels than low-resource regions (e.g. Africa, parts of Asia). This talk will cover recent applications of unsupervised learning and transfer learning methods to remote sensing data in order to achieve higher performance on small quantities of labels than traditional supervised learning.