With increasing deployment of wireless devices also localization gets a consequential boost in industrial and other IoT environments. Localization so far in most cases assumes omnidirectional antennas. This paper investigates a generalized Received Signal Strength (RSS) fingerprinting localization algorithm for antennas with arbitrary radiation patterns in order to include IoT devices with directional antennas in localization.Focus is set on the recursive algorithm which improves performance of the traditional fingerprinting methods. First a machine-learning aided analysis of the overlapping reception areas of the antennas is conducted for quickly determining possible initial positions of the nodes to be localized. Then, for reaching higher resolution, different up-scaling techniques and gradient search is applied. In particular, the localization accuracy vs. computational efficiency is investigated which is an important trade-off for many applications.