In this report, a methodology to automatically detect the presence of fog on the RMI webcam images is tested. For that purpose, a convolutional neural network for image classification was constructed in Python using TensorFlow, a free and open-source software library for machine learning originally developed by Google. A first model was constructed for the webcam images of Mont Rigi, using five distinct visibility classes. The insights gained from this model were then used to construct a set of models for six individual webcams and only three visibility classes ("fog", "no fog", "hard to see"), resulting in a classification accuracy higher than 98% on the validation dataset, except for the webcam of Mont Rigi for which ~96% was reached. A combined model that can be applied to images originating from a random webcam (that was included in the training dataset) still reaches ~98% validation accuracy. The labeling process as well as the exploration of different model settings are discussed in the report. It is concluded that ML based image classification techniques can be used for the efficient detection of fog. The presented model could be further developed as an auxiliary tool to notify the RMI forecasters in real time when fog has been detected on one of the webcams of the RMI network.