DSFormer: A Dual-Scale Cross-Learning Transformer for Visual Place Recognition | IEEE Journals & Magazine | IEEE Xplore