Check out our recent cloudmask processor results in comparison with another machine learning based cloudmask. We specifically selected the worst performing sub-tiles where different kind of errors may be still visible.

S2cloudless is a single-scene cloud detection algorithm which runs single pixel-based classification using different models, such as decision trees, support vector machines (SVM) and neural networks. They claim that even with high performance neural networks, they have considerably slow inference time and high computational complexity. S2cloudless outperforms existing single-scene algorithms Fmask and Sen2Cor. However, S2cloudless focuses only on cloud detection, ignoring cloud shadows. The code is open-source and available here https://github.com/sentinel-hub/sentinel2-cloud-detector .

For comparison purposes with cloud only S2cloudless we mapped KZ prediction of cloud and semi-transparent classes to cloud. Read more about KappaZeta’s cloudmask in our blog: https://kappazeta.ee/blog/free-and-open-ai-based-cloud-mask-for-sentinel-2.

The following examples are tiles from the product S2B_MSIL2A_20200401T093029_N0214_R136_T34UFA_20200401T122148.SAFE