Many applications require knowledge about the position of objects in a room. Popular ways to tackle this issue is to use either vision based sensors, or several communicating beacons placed at known positions which allow beamforming or triangulation methods. However, in some cases, vision is limited due to a lack of light or the presence of airborne obscurants and also the placement of several beacons can be seen as impractical. This paper suggests a method using a monostatic setup where a sensor uses a limited set of known Room Impulse Responses to then accurately estimate its position in that environment using a Regression Convolutional Neural Network. The research is performed using a Finite-difference Time-domain simulation method to generate realistic data and achieves results with an average estimation error of 14, 7 cm.