Joint Extraction of Retinal Vessels and Centerlines Based on Deep Semantics and Multi-Scaled Cross-Task Aggregation | IEEE Journals & Magazine | IEEE Xplore