IBM Research and the Horizon Europe Embed2Scale consortium are inviting the AI community to redefine and advance compression for Earth Observation data.
The exponential growth in satellite imagery means that storing, transferring, and analyzing these vast datasets is becoming increasingly complex and expensive —yet this data is essential for unlocking insights across diverse applications. Modern neural compression techniques have emerged as a promising solution, achieving impressive compression rates while preserving the essential details needed for downstream tasks.
Our 2025 CVPR EARTHVISION Data Challenge invites researchers and AI practitioners to develop and benchmark innovative compression methods, with the winning solution showcased at CVPR EARTHVISION 2025.
Discover more about neural compression and get details on the challenge in our latest Medium Blog post.
#MachineLearning