Lie groups; Feature extraction; Image classification; Image motion analysis; Image recognition; Image representation; Learning (artificial intelligence); 3D classification; Lie group structure; Appropriate Lie group; Deep learning methods; Deep network architecture; High feature dimensionality; Input Lie group; Logarithm mapping layer; Skeleton; Standard 3D human action recognition datasets; Computer architecture; Machine learning; Manifolds; Neural networks; Three-dimensional displays; Transforms