Abstract: Traditional and deep Structure-from-Motion (SfM) methods typically operate under the assumption that the scene is rigid, i.e., the environment is static or consists of a single moving object ...