Visual Place Recognition (VPR) is the task of recognizing what place is represented in a given image. This task, which is being studied for decades, is rich of challenges and margins for improvements. The main purpose of this thesis was to study techniques to improve the scalability of VPR algorithms without penalizing their precision. This work contains many experiments that show which improvement can be achieved by applying similar architectures and what their limitations are. Moreover, a software made up by APIs back-end calls and a user-friendly front-end, was developed to potentially allow every user to perform searches on large scale in just a matter of seconds.