Diagram of the Siamese CNN training phase. During training, triplets of velocity maps are passed through identical CNN encoders with shared weights. Each triplet consists of a reference (Anchor), a matched projection from the same simulated cluster (Positive), and a mismatched projection from a different simulated cluster (Negative). The network learns to produce similar embeddings for the Anchor–Positive pairs and dissimilar embeddings for the Anchor–Negative pairs by minimizing the triplet loss. This process teaches the model to recognize the underlying kinematic structure of clusters, independent of projection and resolution differences.