From the beginning of the 2000’s, cameras are considered as an interesting source of opportunistic meteorological data. This short study deals with the comparison of meteorological visibility between images. A new dataset has been built from publicly available webcam sequences. An original labeling process, based on a mergesort algorithm, allowed us to sort more than 400 webcam sequences with respect to the meteorological visibility. Standard CNN have been trained on these sequences in a basic “learning to compare” framework and tested on independent webcams that are colocalized with visibilimeters. Results on the comparison task are promising. We observe that taking into account the numerous abstention cases improves our predictions.