 The multi-spectral imager on board the Sentinel-2s of the Copernicus program offers optimized bands for the task of classification of clouds, cirrus, snow, shadows, and clear sky areas. Additionally, it provides global coverage, spectral coverage, and repetition rates. Efficient algorithms are necessary to process these large amounts of data. Techniques based on top-of-atmosphere reflectance spectra for single pixel without external data or spatial context offer the greatest potential for parallel data processing and high optimization of processing throughput. Decision trees, random forests, stochastic gradient descent, and support vector machines have been used for this task. The presented algorithms based on the classical Bayesian method have shown excellent classification skill and processing performance.