Figure S6. 2D and 3D Visualizations of facial data distribution. Pairwise distance vectors of 13 facial landmarks from facial images and augmented dataset are visualized in 2D plane using t-Stochastic Neighborhood Embedding (t-SNE) with perplexity 30. The clustered data points of each class imply that the facial expressions can be read by the pairwise distance vectors of the 3D facial images.