All economic sectors must understand, measure and mitigate their contributions to climate change. The aviation sector is no exception and has to reduce its CO2 emissions while also addressing its non- CO2 effects which are responsible for a significant radiative impact on climate. The most important of these effects is due to the formation of contrails and their transformation into induced cirrus. Many studies have focused on detecting contrails onto satellite images because, taken together, meteorological geostationary and sun-synchronous satellites provide a good monitoring of the Earth’s atmosphere, but unfortunately the spatial resolution and temporal sampling of such satellite images are often insufficient to detect contrails right after their formation and attribute a particular contrail to a given flight. The use of ground-based cameras, especially as part of a network, is therefore complementary to satellite imagery and currently represents an important avenue of research for contrail monitoring. In this article we describe a dataset of annotated ground-based hemispheric sky images that can serve as a basis for the training and validation of contrail detection algorithms, in particular those aiming at segmenting contrails using machine learning methods.