Support vector machines (SVMs) and random forest (RF) represent promising developments in machine learning research and have been used widely in the remote-sensing community within past decades. In spite of their popularity for land use and cover classication, SVMs still have some problems. This chapter discusses various issue related to the design of SVMs and compare its performance with RF classiers in terms of classi-cation accuracy, computation cost, various user-dened parameters, and feature selection. Two data sets-one multispectral and other hyperspectralare used to compare the performance of SVMs and RF classiers. Results suggest the usefulness of RF classiers in comparison to SVMs in terms of classication accuracy, computation cost, and feature selection. Results also suggest that such issues as the requirement of a set of user-dened parameters, choice of a suitable multiclass classication approach, and kernel function need attention while using SVMs for land cover classication.