PartFormer: Awakening Latent Diverse Representation from Vision Transformer for Object Re-Identification | Synapse