Remote sensing is entering a new era of time-series analysis. Short revisit times of satellites allow for monitoring of many areas across the globe on a weekly basis. However, there has been little exploration of deep learning techniques to leverage this new temporal dimension at scale. Especially, existing approaches have struggled to combine the power of different sensors to make use of all available information. In addition, large scale high quality change detection benchmarks are rare. 

To stimulate innovation in spatio-temporal machine learning, our ELLIS Munich unit members Xiaoxiang Zhu and Laura Leal-Taixé partnered up with TUM, DLR and Planet labs to organize a unique challenge centered around modeling multi-temporal land cover changes from Planetscope and Sentinel time series data, as part of the EarthVision Workshop at CVPR 2021. This effort is jointly supported by Helmholtz AI, TUM, DLR, Planet labs, BMWi, German Space Agency, International AI future lab 'AI4EO', Munich Data Science Institute and Munich Data Science Research School.

DynamicEarthNet Challenge: http://www.classic.grss-ieee.org/earthvision2021/challenge.html