Self-supervised 3D shape variational autoencoder for latent representation of identity features - 42Papers