It is always crucial to ensure that learning traditional art forms is easily accessible, and thus there is a growing need to have an automated system that helps learners correct themselves. This study focuses on one such art form, Bharatanatyam, an ancient South Indian dance form that conveys its storyline through beautiful movements involving various parts of the body. Here, we focus on “Thattadavu”, which is of 8 types. Thattadavus are leg movements that have various tempos. Using the spatial information a CNN can provide, we combine it with LSTMs to extract the temporal information to create our ConvLSTM model. This model gives out high accuracy and precision in such dance movement recognition tasks since it focuses on the recurrent nature of the video frames. The work also uses a fully novel dataset, which was carefully curated with a proper balance across all the 8 classes.