DSCAFormer: Lightweight Vision Transformer With Dual-Branch Spatial Channel Aggregation | IEEE Journals & Magazine | IEEE Xplore