Expediting Large-Scale Vision Transformer for Dense Prediction Without Fine-Tuning | IEEE Journals & Magazine | IEEE Xplore