The CloudSEN12 project started in Peru with the guidance of Z_GIS and the ISP. This initiative addresses the challenge of processing cloud-contaminated imagery, focusing on tropical regions where existing cloud detection algorithms frequently underperform. We quickly identified that this shortcoming was due to the absence of high-quality labels on this region. To address this issue, we decided to create a globally diverse dataset with a strong focus on quality to improve SOTA cloud detection capabilities. After the dataset’s release in December 2022 and receiving valuable feedback, the team was motivated to improve further. This led to CloudSEN12+, where, with the acquired knowledge, we refined the dataset ensuring maximum trustworthiness.

Mission:

Provides high-quality, open cloud detection labels for Sentinel-2, ensuring fair global data distribution.

Vision:

Our aim is to become the most reliable provider of cloud semantic annotations across multiple optical remote sensing sensors. We strive to provide researchers and industries with the necessary models and datasets to perform accurate cloud screening.