During the past decade or so, standard photogrammetric practice has experienced an impressive influence by computer vision algorithms, which have led to its automation and democratization. Although this radical change has obvious advantages, it is also escorted by some severe disadvantages. Potential users are faced with numerous parameter choices about which they have no expert knowledge and consequently, the results are not always ideal, to say the least. In this chapter, we try to pinpoint the lurking dangers and help non-expert users to take knowledgeable decisions and be able to evaluate their results. The few—unavoidable—theoretical aspects are escorted by rich illustrated examples, thus leading to practical guidelines for achieving optimum results.