Orientation-invariant streamline dissimilarity embedding, MiniBatch K-Means clustering, and medoid extraction