Perception

Unsupervised autoencoder learns spatial context from trajectory data for annotation.