This article discusses on the basic properties useful for characterizing various kinds of atmospheres possible to be created by the students during a class, as well as the observation features to be obtained from the video of the class for estimating the degrees of the atmospheres for the basic properties, aiming to recognize the classroom atmosphere from its video for reviewing the class for further improvement. We propose a pair of basic properties, which we call “interest–disinterest” and “liberal–restrained”, by applying Russell's circumplex model of affect to characterization of atmospheres, and show that the rate of the students looking forward and the amount of frame difference of the video respectively have good correlations with the degree of the atmospheres characterized by those two basic properties from the result of an experiment using videos of a seminar in a university class.