In this work, we aim to perform auditory source localization in images. We want not only to estimate which objects in the scene are making the annoying sound but also to infer how their sounds interact with the environment. To achieve that, we propose a methodology that uses a classification task and Class Activation Mapping for doing the PA localization, and a pipeline composed of a depth estimation neural network and a sound propagation library for estimating the PA propagation on the environment. Our experiments show that our method achieves satisfactory auditory source localization, and it can generate a map that represents how the annoying sound is propagating in the scene.