It has recently been popular to use neural network based classifiers in remote sensing. The reported results are usually basedon a very limited data set. This paper concerns the experiences from the cloud classification scheme. The experiences are based on a data set of several hundred images. The classification is based on Self-Organizing Maps (SOM), which are fine-tuned by the Learning Vector Quantization (LVQ). The classification is done in two phases, the clouds are first screened and then classified. The classifier is capable of classifying satellite images taken round the year, during day and night. The classifier is fully automatic, and it can be adapted to changing situations with new examples. The benefits of this approach are rapid prototyping, adaptivity, and in high degree unsupervised learning.