Multi-Resolution Networks for Semantic Segmentation in Whole Slide Images