MSViT: Training Multiscale Vision Transformers for Image Retrieval | IEEE Journals & Magazine | IEEE Xplore