Abstract: Aiming at the challenges of Vision Transformer (ViT) model in terms of computational cost, an adaptive pruning strategy for tokens based on Closeness-Centrality is innovatively proposed. The ...