This paper presents an effective computer vision methodology for monitoring displacements in multi-story buildings during shake-table testing, as a convenient alternative to the installation of conventional contact sensors. in the proposed system, in its most simple spatial configuration, exploits only two video cameras. One camera is installed on the roof of the building pointing downward and simultaneously measures the absolute and relative horizontal displacements of the targets installed at the different floor levels. A second camera considered as stationary reference is installed at the ground level outside of the shake-table and it acquires the displacement of the roof, tracking a target installed in the proximity of the internal camera. In this way, it is possible to have: i) a redundant measurement of the roof displacement; ii) the opportunity to compensate for possible sources of noise induced in the internal video camera on the roof by the shaking. The first application of the proposed vision-based methodology occurred during the shake-table testing of a full-scale, six-story mass timber building tested on the 6-DOF outdoor shake-table (LHPOST6) at the University of California San Diego. Based on the results achieved a critical discussion of the vision-based methodology performance is presented.