Visual navigation of robots is a trending topic in the research field. This paper presents a target-driven navigation method that uses Transformers to achieve optimal environmental perception. With this perception, a deep reinforcement learning (DRL) agent based on the Soft Actor Critic (SAC) algorithm can take optimal actions to reach the target. Our method uses a depth image with a field of view (FOV) limited to about 66°, allowing for more precise perception and improved generalization capability of our model. The results show superior performance compared to the model using a Fisheye image with a 220-degree FOV. The source code is available at the following link: https://github.com/REGRAGUIahmed/vis2nav/tree/master , along with a demonstration at: https://youtu.be/Bv8yX3H_R3w .